Benchmarks.bio - scBench-Long: leaderboard
scBench-Long (Benchmarks.bio) — long-form single-cell biology question answering scored against expert references.
Metric: Pass Rate (%). Source: benchmarks.bio. Saturation forecast: Around May 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 41.27 | #12 |
| 2 | GPT-5.6 Sol | 38.1 | #19 |
| 3 | Claude Opus 4.8 | 25.4 | #37 |
| 4 | GPT-5.6 Terra | 23.81 | #41 |
| 5 | Gemini 3.5 Flash | 22.22 | #60 |
| 6 | GPT-5.6 Luna | 19.05 | #68 |
| 7 | Gemini 3.1 Pro (Preview) | 17.46 | #58 |
| 8 | GPT-5.5 | 15.87 | #31 |
| 9 | Claude Sonnet 5 | 14.29 | #57 |
| 10 | Kimi K3 | 12.7 | #30 |
| 11 | Grok 4.5 | 12.7 | #48 |
| 12 | Claude Opus 4.7 | 11.11 | #49 |
| 13 | GPT-5.4 | 9.52 | #79 |
| 14 | Claude Opus 4.6 | 9.52 | #63 |
| 15 | GLM-5.2 | 9.52 | #73 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=benchmarks-bio-scbench-long · How It Works · Data refreshed daily, snapshot 2026-10-20.