Skip to content

feat(inference): open Qwen3.5 Agentic coding on FP4+FP8 (same as GLM-5.2/5.3) #1361

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
functionstackx merged 5 commits into from
Oct 16, 2026
Merged

feat(inference): open Qwen3.5 Agentic coding on FP4+FP8 (same as GLM-5.2/5.3) #1361

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
f44b88c
feat(inference): open Qwen3.5 Agentic coding on FP4+FP8 (same as GLM-…
kedarpotdar-nv Oct 15, 2026
9039faf
test: drop model-specific Qwen3.5 default-precision test
kedarpotdar-nv Oct 15, 2026
8ad9ae9
test: drop the stale Qwen3.5 no-override assertion
kedarpotdar-nv Oct 15, 2026
714fdb0
test: remove getModelDefaultPrecisions tests per review
kedarpotdar-nv Oct 16, 2026
7dd75a4
Merge branch 'master' into kedar/qwen35-agentic-fp4-fp8
kedarpotdar-nv Oct 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 0 additions & 15 deletions packages/app/src/lib/catalog/data-mappings.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,6 @@ import {
isSequenceDeprecated,
isSequenceDeprecatedForModel,
isSequenceHiddenForModel,
getModelDefaultPrecisions,
getSequenceCategoryForModel,
isBestPerSkuDefaultOff,
showsTcoBasisSelector,
Expand Down Expand Up @@ -480,17 +479,3 @@ describe('isSequenceHiddenForModel', () => {
expect(isSequenceHiddenForModel(Model.MiniMax_M3, Sequence.EightK_OneK)).toBe(false);
});
});

describe('getModelDefaultPrecisions', () => {
it('opens GLM-5.2 / GLM-5.3 Agentic coding on FP4 and FP8', () => {
expect(getModelDefaultPrecisions(Model.GLM_5_2, Sequence.AgenticTraces)).toEqual([
Precision.FP4,
Precision.FP8,
]);
});

it('leaves other scenarios and models on the auto default', () => {
expect(getModelDefaultPrecisions(Model.GLM_5_2, Sequence.EightK_OneK)).toBeUndefined();
expect(getModelDefaultPrecisions(Model.Qwen3_5, Sequence.AgenticTraces)).toBeUndefined();
Comment thread
kedarpotdar-nv marked this conversation as resolved.
Resolved
});
});
2 changes: 2 additions & 0 deletions packages/app/src/lib/catalog/data-mappings.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -686,13 +686,15 @@ export function getPrecisionLabel(precision: Precision): string {
* manual toggle still wins; precisions the scenario lacks are dropped.
*
* GLM-5.2 / GLM-5.3 Agentic coding: FP4 and FP8 both carry full fleets.
* Qwen3.5 397B Agentic coding: same; FP4 and FP8 both carry full fleets.
*
* Declared after `Precision`: TypeScript enums are initialised in module order.
*/
const MODEL_DEFAULT_PRECISIONS: Partial<
Record<Model, Partial<Record<Sequence, readonly Precision[]>>>
> = {
[Model.GLM_5_2]: { [Sequence.AgenticTraces]: [Precision.FP4, Precision.FP8] },
[Model.Qwen3_5]: { [Sequence.AgenticTraces]: [Precision.FP4, Precision.FP8] },
};

/** The fixed default precision set for this model and scenario, if any. */
Expand Down
Loading