Benchmarks · agentic
BioMysteryBench
Anthropic's benchmark of whether AI agents can solve real-world bioinformatics mysteries from raw data
21 models · 21 results · updated 2026-10-24 · vals.ai ↗
Leaderboard
Sonnet 5.5 max
More from this benchmark
1 published measuresTime per task29 results · lower is better
All results
21 rows| # | Model | Effort | BioMysteryBench | Subscription $ / task | API $ / task | Time | Harness |
|---|---|---|---|---|---|---|---|
| 1 | Sonnet 5.5 | max | 81.1% | $0.057 | $3.74 | 18 min | Terminus 2 |
| 2 | GPT-6.1 Sol | max | 79.6% | $0.043 | $0.37 | 8.8 min | Terminus 2 |
| 3 | Opus 5.5 | max | 79.3% | $0.055 | $3.64 | 12 min | Terminus 2 |
| 4 | Opus 5 | max | 79.3% | $0.057 | $3.29 | 37 min | Terminus 2 |
| 5 | GPT-6 Astra | max | 79.3% | $0.15 | $1.72 | 4.8 min | Terminus 2 |
| 6 | Gemini 4 Argon | high | 76.3% | $2.03 | $5.71 | 40 min | Terminus 2 |
| 7 | GPT-6 Sol | max | 74.8% | $0.041 | $0.66 | 5.4 min | Terminus 2 |
| 8 | Kimi K3 | max | 72.6% | $0.082 | $1.06 | 4.9 min | Terminus 2 |
| 9 | Grok 4.6 | high | 72.2% | $0.13 | $1.23 | 14 min | Terminus 2 |
| 10 | GPT-5.6 Sol | max | 71.1% | $0.11 | $1.93 | 18 min | Terminus 2 |
| 11 | Haiku 5.5 | max | 70.4% | $0.011 | $0.51 | 13 min | Terminus 2 |
| 12 | Grok 4.7 | xhigh | 69.3% | $0.23 | $2.49 | 12 min | Terminus 2 |
| 13 | MiMo V2.6 Flash | high | 69.3% | free | $0.050 | 28 min | Terminus 2 |
| 14 | DeepSeek V4.1 Flash | high | 67.8% | $0.015 | $0.23 | 15 min | Terminus 2 |
| 15 | Mistral Large 4 | high | 67.0% | $2.07 | $2.07 | 26 min | Terminus 2 |
| 16 | Muse Spark 1.2 | xhigh | 64.8% | $0.50 | $0.99 | 24 min | Terminus 2 |
| 17 | DeepSeek V4 Flash 0731 | high | 64.4% | $0.039 | $0.039 | 27 min | Terminus 2 |
| 18 | Gemini 3.8 Flash | high | 62.2% | $0.063 | $1.55 | 5.4 min | Terminus 2 |
| 19 | GPT-5.6 Luna | max | 61.5% | $0.008 | $0.089 | 19 min | Terminus 2 |
| 20 | GPT-6 Luna | max | 61.5% | $0.008 | $0.053 | 8.7 min | Terminus 2 |
| 21 | Gemini 3.6 Flash | high | 58.5% | $0.039 | $1.25 | 13 min | Terminus 2 |