Skip to content

Feature/mistral thinking fix#174

Merged
solderzzc merged 3 commits intodevelopfrom
feature/mistral-thinking-fix
Mar 21, 2026
Merged

Feature/mistral thinking fix#174
solderzzc merged 3 commits intodevelopfrom
feature/mistral-thinking-fix

Conversation

@solderzzc
Copy link
Member

No description provided.

solderzzc and others added 3 commits March 19, 2026 22:45
…el leaderboard

- Switch to 1-column onecolumn IEEEtran format
- Update hardware: M3 24GB → M5 Pro 64GB (18 cores, 30 GPU cores)
- Expand model coverage: 3 configs → 7 models
  - Local: Qwen3.5 9B/27B/35B-MoE/122B-MoE (GGUF)
  - Cloud: GPT-5.4, GPT-5.4-mini, GPT-5.4-nano
- Add TTFT + tok/s inference performance table
- Add cross-model event deduplication breakdown table (8 tests × 7 models)
- Update abstract, contributions, results, failure taxonomy, deployment matrix
- Key finding: Qwen3.5-9B scores 93.8% (only 4.1pp below GPT-5.4), and
  Qwen3.5-35B-MoE has lower TTFT (435ms) than all cloud models tested
…anity check

- Add MODEL_FAMILIES config table with per-model API params and server flags
- Add getModelApiParams() helper to inject reasoning_effort:none for Mistral
- Add delta.thinking fallback in streaming loop to capture thinking tokens
- Add streaming sanity check before benchmark run (detects empty-token loops)
- Add test-model-config.cjs with 17 unit tests for model detection logic
@solderzzc solderzzc merged commit a304253 into develop Mar 21, 2026
1 check passed
@solderzzc solderzzc deleted the feature/mistral-thinking-fix branch March 21, 2026 06:57
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.

1 participant