Modality: llm_chat · full deep dive — every ranked model, test result, and artifact.
Ranked by confidence-adjusted score (single/zero-sample, non-curated scores floored; curated empirical scores trusted as-is).
| # | Model | Provider | Adj. Score | Raw | Evidence |
|---|---|---|---|---|---|
| 1 | gpt-5.5 | codex | 0.846 | 0.931 | n=50 |
| 2 | gpt-5.5::think=deep | codex | 0.525 | 0.963 | n=6 |
| 3 | gpt-5.5::think=fast | codex | 0.525 | 0.963 | n=6 |
| 4 | gemini-2.5-flash | google_gemini | 0.455 | 0.500 | n=50 |
| 5 | claude-haiku-4-5-20251001 | anthropic | 0.455 | 0.500 | n=50 |
| 6 | claude-haiku-4-5-20251001 | claude_code | 0.196 | 0.216 | n=50 |
| 7 | gpt-5.4-mini | codex | 0.150 | 1.000 | provisional (n≤1) |
No benchmark outputs recorded for this niche yet.