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 | claude-sonnet-4-6 | anthropic | 0.920 | 0.920 | curated |
| 2 | claude-opus-4-6 | anthropic | 0.840 | 0.840 | curated |
| 3 | deepseek-reasoner | deepseek | 0.700 | 0.700 | curated |
No benchmark outputs recorded for this niche yet.