FrontierScience

A benchmark for research-level scientific reasoning, designed to separate frontier models on difficult science tasks that mix domain knowledge with deep reasoning.

ModelsModels with a result.2
Top resultBest result on this benchmark.—
Top-3 spreadPoints from first to third.—
YearYear of release.—

Results

2 results
#ModelResultThe score from the source.
—GPT-5.4 Pro36.7%