Skip to content

Update OpenAI models to the GPT-6 lineup - #3

Merged
Gvo87 merged 1 commit into
mainfrom
update-openai-models
Oct 16, 2026
Merged

Gvo87 merged 1 commit into
mainfrom
update-openai-models

Conversation

@Gvo87

@Gvo87 Gvo87 commented Oct 15, 2026

Copy link
Copy Markdown
Owner

Summary

  • Current models in Settings: gpt-6-luna, gpt-6-astra, gpt-6.1-sol, gpt-6-sol, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna. Removed models OpenAI has deprecated or shut down: gpt-4o, gpt-4.1*, gpt-5/-mini/-nano, gpt-5-chat-latest. The combo box is still editable, so any model name can be typed in.
  • New default gpt-6-luna, OpenAI's recommended replacement for gpt-4o.
  • GPT-6 requests no longer fail with HTTP 400. Every GPT model from GPT-5 onward is now treated as a reasoning model: requests send max_completion_tokens and no temperature. Previously only gpt-5* and the o-series were, so GPT-6 requests sent max_tokens and temperature: 0.7, which those models reject.
  • Saved retired models switch to gpt-6-luna once. A new SettingsVersion registry value records that the switch ran, so if someone picks an older model again later, that choice is kept.
  • The model checks moved into src/ModelParams.{h,cpp} and are covered by new unit tests.

Model IDs and deprecation dates are from OpenAI's models and deprecations pages.

Testing

  • Unit tests pass (11/11, including 4 new model-check cases).
  • Plugin DLL builds (Release x64).
  • Installed in Notepad++ over a saved gpt-5-mini setting. Settings showed gpt-6-luna, Test Connection succeeded with gpt-6-luna, and the registry was saved as Model = gpt-6-luna, SettingsVersion = 1.
  • Not tested by hand: choosing an older model again afterwards and confirming it is kept after a restart.

🤖 Generated with Claude Code

- Offer current models in Settings (gpt-6-luna, gpt-6-astra, gpt-6.1-sol,
  gpt-6-sol, gpt-5.6-sol/terra/luna) and drop deprecated or shut-down ones
  (gpt-4o, gpt-4.1*, gpt-5/-mini/-nano, gpt-5-chat-latest).
- Default to gpt-6-luna, OpenAI's recommended replacement for gpt-4o.
- Treat every GPT model from GPT-5 onward as a reasoning model so GPT-6
  requests send max_completion_tokens and no temperature instead of
  failing with HTTP 400.
- Switch saved retired models to gpt-6-luna once, tracked by a new
  SettingsVersion registry value so a later explicit choice is kept.
- Move the model checks into ModelParams and cover them with unit tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🔵 Needs a closer look

Model availability, retirement assumptions, and API compatibility remain unverified, and persistence of reselected older models needs confirmation.

0 open findings

What changed in this PR

Updates the plugin’s model choices and default, extends reasoning-model request handling, and introduces versioned migration of saved model settings.

Changes:

  • Replaces preset models and sets gpt-6-luna as the default.
  • Extracts model classification helpers and adds unit tests.
  • Adds a one-time migration for models classified as retired.
File Description
tests/​CoreTests.cpp Adds model classification tests.
src/​SettingsDialog.cpp Updates selectable model presets.
src/​ModelParams.h Declares model classification helpers.
src/​ModelParams.cpp Implements reasoning and retirement checks.
src/​LLMService.cpp Updates the default and adds versioned settings migration.
LLMAssistant.vcxproj Includes the new helpers in the plugin build.
CMakeLists.txt Includes model helpers in the test target.

🧠 Review effort: Balanced


💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

@Gvo87
Gvo87 merged commit 6774c81 into main Oct 16, 2026
2 checks passed
@Gvo87 Gvo87 mentioned this pull request Oct 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants