ScBench V1 Clean Reruns
Science
Clean-v1 reruns on 195 single-cell analysis evaluations. Formerly V2, now V1; macro-mean pass rates on updated evaluation sets remain separate from the original cohort.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 63.9 |
| GPT-6.1 Sol | 63.6 |
| GPT-6 Sol | 61.5 |
| Claude Sonnet 5.5 | 60.3 |
| GPT-5.6 Sol | 59.8 |
| Grok 4.6 | 59.1 |
| Grok 4.7 | 58.5 |
| Claude Opus 5 | 58.3 |
| Opus 4.8 | 57.1 |
| GPT-5.5 | 57.1 |
| Claude Opus 5.5 | 56.4 |
| Sonnet 5 | 55.6 |
| DeepSeek V4.1 Flash | 54.7 |
| GPT-6 Luna | 53.9 |
| Opus 4.7 | 52.5 |
| Gemini 3.7 Flash | 50.4 |
| GPT-5.6 Luna | 50.3 |
| Kimi K3 | 47.5 |
| Sonnet 4.6 | 41.7 |
| Gemini 3.8 Flash | 34.0 |
| Nemotron 3.5 Lightning 30B A3B (precision unspecified) | 22.9 |