Benchmarks / Reasoning
benchmarks.bio Expected Performance
Mean IRT-predicted pass rate across automatically discovered evaluations, excluding refusal. Uses the best mapped harness per model and evaluation and requires observed coverage in every domain; the aggregate counts as one benchmark. Score: IRT expected performance across evaluations. Collected by Jake Boggs' benchmark dashboard.
benchmarks.bio Expected Performance
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GPT-6 Astra | max | 53.1% |
| 2 | Opus 5 | max | 50.8% |
| 3 | Grok 4.6 | max | 47.8% |
| 4 | Opus 4.8 | max | 46.0% |
| 5 | GPT-5.6 Sol | max | 43.4% |
| 6 | Gemini 3.7 Flash | max | 43.2% |
| 7 | DeepSeek V4.1 Flash | max | 42.9% |
| 8 | Gemini 3.5 Flash | xhigh | 42.2% |
| 9 | Sonnet 5 | max | 42.0% |
| 10 | Grok 4.5 | max | 41.7% |
| 11 | GPT-5.6 Terra | max | 41.5% |
| 12 | GPT-5.5 | xhigh | 41.2% |
| 13 | GPT-5.6 Luna | max | 32.9% |