Benchmarks
347 sources · 1,258 models × 3,038 benchmarks · 158,912 scoresBioMysteryBench tests whether agents can interleave computational analysis with biological reasoning to solve difficult challenges from unprocessed scientific datasets.
No signal or difficulty estimate. Held out of the fit by design.
This is a benchmark-family page. Scores are recorded on the component benchmarks below rather than directly on the parent, because the components use different score scales and directions.
| Metric | Models | Leader | Direction | Date |
|---|---|---|---|---|
| BioMysteryBench - Human Difficult | 9 | Gemini 3.8 Flash56.5 | higher is better | 2026-07-24 |
| BioMysteryBench - Human Solvable | 9 | Claude Opus 590.1 | higher is better | 2026-07-24 |
| BioMysteryBench - Human Difficult (original problem set) | 1 | Claude Fable 546.1 | higher is better | — |
| BioMysteryBench - Human Solvable (original problem set) | 1 | Claude Fable 583.9 | higher is better | — |