EpiBench
Science · 2026-06-11
Epigenomics benchmark covering ATAC-seq, ChIP-seq, methylation, and CUT&RUN/CUT&Tag analysis tasks on real data.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.5 | 45.0 |
| GPT-6 Astra | 44.6 |
| DeepSeek V4.1 Flash | 42.1 |
| GPT-5.6 Sol | 42.1 |
| GPT-5.6 Terra | 39.9 |
| Opus 4.8 | 39.0 |
| GPT-5.4 | 39.0 |
| Sonnet 5 | 37.7 |
| Grok 4.6 | 37.4 |
| GPT-5.6 Luna | 36.5 |
| Claude Opus 5 | 35.2 |
| Opus 4.7 | 31.4 |
| Opus 4.6 | 30.5 |
| Gemini 3.5 Flash | 28.9 |
| Grok 4.5 | 28.9 |
| Gemini 3.1 Pro Preview | 27.4 |
| Kimi K3 | 26.7 |
| Kimi K2.6 | 24.5 |
| Sonnet 4.6 | 22.6 |
| Grok 4.20 | 22.3 |
| Grok 4.3 | 12.6 |