BioMysteryBench (Vals AI) leaderboard
- Claude Sonnet 5.5 leads
- 81.1%
- models measured
- 26
- lead over #2
- +1.5 pts
- median
- 69.3%
Leader on BioMysteryBench (Vals AI) · read October 22
On October 22, 2026, 26 of our catalog's models had a BioMysteryBench (Vals AI) result, and Anthropic Claude Sonnet 5.5 led with 81.1%, ahead of OpenAI GPT-6.1 Sol at 79.6%.
End-to-end bioinformatics investigations by an agent. Share solved, as Vals AI publishes it.
Role in the index
BioMysteryBench (Vals AI) is shown as published and does not feed any of our indexes. Read the methodology
| # | Model | Result | Cost per test | Duration, as printed | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5Anthropic | 81.1% | $3.74 | 18m 06s | 87.3 | $10.00 | |
| 2 | GPT-6.1 SolOpenAI | 79.6% | $0.37 | 8m 51s | 82.4 | $10.00 | |
| 3 | Claude Opus 5Anthropic | 79.3% | $3.29 | 36m 36s | 83.0 | $25.00 | |
| 4 | Claude Opus 5.5Anthropic | 79.3% | $3.64 | 12m 19s | 88.8 | $20.00 | |
| 5 | GPT-6 AstraOpenAI | 79.3% | $1.72 | 4m 46s | 85.1 | $50.00 | |
| 6 | Gemini 4 ArgonGoogle | 76.3% | $5.71 | 39m 57s | 81.0 | $10.00 | |
| 7 | GPT-6 SolOpenAI | 74.8% | $0.66 | 5m 24s | 76.0 | $10.00 | |
| 8 | Kimi K3Moonshot AI | 72.6% | $1.06 | 4m 55s | 62.3 | $15.00 | |
| 9 | Grok 4.6xAI | 72.2% | $1.23 | 14m 18s | 65.7 | $6.00 | |
| 10 | GPT-5.6 SolOpenAI | 71.1% | $1.93 | 18m 18s | 73.5 | $20.00 | |
| 11 | Claude Haiku 5.5Anthropic | 70.4% | $0.51 | 12m 32s | 70.7 | $0.50 | |
| 12 | Grok 4.7xAI | 69.3% | $2.49 | 11m 37s | 66.7 | $6.00 | |
| 13 | Hy4 previewTencent | 69.3% | $0.36 | 23m 30s | 44.3 | $2.501 | |
| 14 | MiMo-V2.6-FlashXiaomi | 69.3% | $0.05 | 28m 28s | 54.7 | $0.28 | |
| 15 | DeepSeek V4.1 FlashDeepSeek | 67.8% | $0.23 | 14m 50s | 62.5 | $0.84 | |
| 16 | Step 5 PreviewStepFun | 67.4% | $0.49 | 21m 35s | 61.1 | $2.70 | |
| 17 | Ember-1Fireworks | 67.0% | $0.91 | 9m 26s | 52.6 | $15.00 | |
| 18 | Muse WonderMeta | 67.0% | $3.23 | 18m 02s | 71.9 | $20.00 | |
| 19 | Mistral Large 4Mistral AI | 67.0% | $2.07 | 25m 32s | 56.4 | $4.18 | |
| 20 | Muse Spark 1.2Meta | 64.8% | $0.99 | 23m 37s | 55.7 | $4.25 | |
| 21 | DeepSeek V4 Flash 0731DeepSeek | 64.4% | $0.04 | 27m 03s | 51.8 | $0.396 | |
| 22 | Gemini 3.8 FlashGoogle | 62.2% | $1.55 | 5m 27s | 61.6 | $3.75 | |
| 23 | GPT-5.6 LunaOpenAI | 61.5% | $0.09 | 18m 31s | 58.0 | $1.20 | |
| 24 | GPT-6 LunaOpenAI | 61.5% | $0.05 | 8m 44s | 57.3 | $0.50 | |
| 25 | Gemini 3.1 Pro PreviewGoogle | 60.0% | $0.71 | 4m 12s | 32.3 | $12.00 | |
| 26 | Gemini 3.6 FlashGoogle | 58.5% | $1.25 | 12m 42s | 62.5 | $3.75 |
Only models in our catalog are listed, and each appears once with its best published run. 137 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out. Cost per test is what Vals AI reports one test cost on the run behind the result, in US dollars. Every other column is a figure the board publishes for the same run.