BioMysteryBench Human-Difficult: leaderboard
Anthropic BioMysteryBench slice covering 23 real-world bioinformatics tasks no human benchmarker solved after QC, evaluated by average accuracy across five trials per problem.
Metric: Accuracy (mixed sources). Source: huggingface.co. Source: benchmarklist.com. Saturation forecast: Around February 2028. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.8 Flash | 56.5 | #21 |
| 2 | Claude Opus 5 | 49.4 | #12 |
| 3 | GPT-5.6 Terra | 49.4 | #41 |
| 4 | Claude Mythos 5 | 46.1 | #2 |
| 5 | GPT-5.6 Sol | 44.7 | #19 |
| 6 | Gemini 3.7 Flash | 43.5 | #23 |
| 7 | Gemini 3.6 Flash | 41.2 | #51 |
| 8 | Claude Opus 4.8 | 40 | #37 |
| 9 | Claude Sonnet 5 | 34.1 | #57 |
| 10 | Claude Mythos Preview | 29.6 | #4 |
| 11 | Claude Opus 4.7 | 27 | #49 |
| 12 | Claude Opus 4.6 | 23.5 | #63 |
| 13 | Claude Sonnet 4.6 | 19.1 | #90 |
| 14 | Claude Haiku 4.5 | 5.2 | #280 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=biomysterybench-human-difficult · How It Works · Data refreshed daily, snapshot 2026-10-20.