MatFlowBench Experimental Pass@4
Science · 2026-10-09
Share of the 20 experimental MatFlowBench tasks that a model passes in at least one of its four rollouts.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 80.0 |
| Claude Sonnet 5.5 | 75.0 |
| Claude Opus 5.5 | 65.0 |
| Gemini 3.8 Flash | 65.0 |
| GPT-6.1 Sol | 65.0 |
| Muse Spark 1.3 | 60.0 |
| Grok 4.7 | 55.0 |
| Hy4 Preview | 55.0 |
| Kimi K3 | 45.0 |