Atlas

Benchmarks

BioMysteryBench

BioMysteryBench tests whether agents can interleave computational analysis with biological reasoning to solve difficult challenges from unprocessed scientific datasets.

Models
0
Top score
—
Leader
—
Updated
—
Resolution
—
Saturation
—

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.

MetricModelsLeaderDirectionDate
BioMysteryBench - Human Difficult9Gemini 3.8 Flash56.5higher is better2026-07-24
BioMysteryBench - Human Solvable9Claude Opus 590.1higher is better2026-07-24
BioMysteryBench - Human Difficult (original problem set)1Claude Fable 546.1higher is better—
BioMysteryBench - Human Solvable (original problem set)1Claude Fable 583.9higher is better—
Loading Atlas data…