BioMysteryBench
End-to-end bioinformatics investigations from raw data to a conclusion. Run by Vals AI.
As of 19 Sept 2026, GPT-6 Astra leads BioMysteryBench on BenchLeader with 79.3%, ahead of Claude Opus 5 at 79.3%, across 11 model configurations with a published result.
- Published by
- Vals AI
- Category
- Knowledge
- Index weight
- Reference only
- Models
- 11
- Data as of
- 19 Sept 2026
Vals AI (vals.ai).
What the test looks like
Each task hands the model a bioinformatics question and the data behind it, and expects the full analysis a research scientist would run.
How it is scored
Percent of investigations reaching the correct conclusion, graded by Vals AI.
What to keep in mind
Small task set and long tasks, so scores move with single results.
- 1GPT-6 Astra (max)79.3%
- 2Claude Opus 579.3%
- 3Grok 4.6 (high)72.2%
- 4Kimi K3 (max)71.5%
- 5GPT-5.6 Sol (max)71.1%
- 6DeepSeek V4.1 Flash (high)67.8%
- 7Muse Spark 1.2 (xhigh)64.8%
- 8DeepSeek V4 Flash 0731 (high)64.4%
- 9Gemini 3.8 Flash (high)62.2%
- 10GPT-5.6 Luna (max)61.5%
- 11Gemini 3.6 Flash (high)58.5%
11 of 11
| # | ||||
|---|---|---|---|---|
| 1 |  GPT-6 AstramaxOpenAI | 79.3% | 71.8 | 2026-09-16 |
| 2 |  Claude Opus 5Anthropic | 79.3% | 66.7 | 2026-09-16 |
| 3 |  Grok 4.6highxAI | 72.2% | 64.5 | 2026-09-16 |
| 4 |  Kimi K3maxMoonshot AIopen ↗ | 71.5% | 67.2 | 2026-09-16 |
| 5 | GPT-5.6 SolmaxOpenAI | 71.1% | 68.8 | 2026-09-16 |
| 6 |  DeepSeek V4.1 FlashhighDeepSeekopen ↗ | 67.8% | – | 2026-09-16 |
| 7 |  Muse Spark 1.2xhighMeta | 64.8% | 64.1 | 2026-09-16 |
| 8 | DeepSeek V4 Flash 0731highDeepSeekopen ↗ | 64.4% | 55.0 | 2026-09-16 |
| 9 |  Gemini 3.8 FlashhighGoogle | 62.2% | 64.5 | 2026-09-16 |
| 10 | GPT-5.6 LunamaxOpenAI | 61.5% | 60.2 | 2026-09-16 |
| 11 | Gemini 3.6 FlashhighGoogle | 58.5% | 61.7 | 2026-09-16 |
Cite as: BenchLeader, “BioMysteryBench leaderboard”, https://www.benchleader.com/benchmarks/vals_biomysterybench, data as of 19 Sept 2026.
BioMysteryBench: questions
- What does BioMysteryBench measure?
- Each task hands the model a bioinformatics question and the data behind it, and expects the full analysis a research scientist would run. Scores are reported in percent of tasks solved; higher is better.
- Which AI model leads BioMysteryBench?
- GPT-6 Astra leads BioMysteryBench with 79.3% as of 19 Sept 2026, ahead of Claude Opus 5 at 79.3%.
- How many models have BioMysteryBench results?
- 11 model configurations have a BioMysteryBench result on BenchLeader, all taken from Vals AI.
- Who runs BioMysteryBench and how often is it updated?
- BioMysteryBench is published by Vals AI. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does BioMysteryBench count toward the BenchLeader Index?
- No. BioMysteryBench is shown for reference but left out of the composite index.