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 (self-reported). Source: benchmarklist.com. Status: years away from saturation. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.8 Flash | 56.5 |
| 2 | Claude Opus 5 | 49.4 |
| 3 | GPT-5.6 Terra | 49.4 |
| 4 | Claude Mythos 5 | 46.1 |
| 5 | GPT-5.6 Sol | 44.7 |
| 6 | Gemini 3.7 Flash | 43.5 |
| 7 | Gemini 3.6 Flash | 41.2 |
| 8 | Claude Opus 4.8 | 40 |
| 9 | Claude Sonnet 5 | 34.1 |
| 10 | Claude Mythos Preview | 29.6 |
| 11 | Claude Opus 4.7 | 27 |
| 12 | Claude Opus 4.6 | 23.5 |
| 13 | Claude Sonnet 4.6 | 19.1 |
| 14 | Claude Haiku 4.5 | 5.2 |
Interactive version: theaggregate.ai/benchmark?slug=biomysterybench-human-difficult · How It Works · Data refreshed daily, snapshot 2026-09-05.