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.
26 models · 26 results · data as of 2026-11-06 · benchmarks.bio ↗
Leaderboard
Opus 5.5 max
All results
26 rows| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Opus 5.5 | max | 54.3% |
| 2 | Sonnet 5.5 | max | 50.6% |
| 3 | Gemini 4 Argon | max | 47.7% |
| 4 | Fable 5.1 | max | 47.2% |
| 5 | Opus 5 | max | 46.9% |
| 6 | GPT-6 Astra | max | 45.8% |
| 7 | GPT-6.1 Sol | max | 44.7% |
| 8 | Grok 4.7 | max | 41.7% |
| 9 | Grok 4.6 | max | 41.5% |
| 10 | GPT-6 Sol | max | 40.7% |
| 11 | Opus 4.8 | max | 39.3% |
| 12 | GPT-5.6 Sol | max | 38.1% |
| 13 | GPT-6.1 Terra | max | 37.5% |
| 14 | Gemini 3.7 Flash | max | 35.8% |
| 15 | DeepSeek V4.1 Flash | max | 35.3% |
| 16 | GPT-5.5 | xhigh | 35.1% |
| 17 | Sonnet 5 | max | 34.6% |
| 18 | Gemini 3.8 Flash | max | 34.1% |
| 19 | GPT-6 Luna | max | 31.6% |
| 20 | Kimi K3 | max | 30.8% |
| 21 | Mistral Large 4.0 | high | 29.0% |
| 22 | GLM 5.3 | max | 27.9% |
| 23 | GPT-5.6 Luna | max | 27.8% |
| 24 | Sonnet 4.6 | max | 26.8% |
| 25 | Nemotron 3 Ultra | max | 19.4% |
| 26 | Nemotron 3.5 Lightning | 11.0% |