GPQA

Unknown

448 expert-level questions in biology, physics, and chemistry. Designed to be unsearchable.

Benchmark Stats

Models17
Papers17
Metrics1

SOTA History

Not enough data to show trend.

accuracy

accuracy

Higher is better

RankModelSourceScoreYearPaper
1o3Editorial82.82026Source
2o4-miniEditorial77.62026Source
3o1Editorial75.72026Source
4o3-miniEditorial74.92026Source
5o1-previewEditorial73.32026Source
6gpt-45-previewEditorial69.52026Source
7gpt-41Editorial66.32026Source
8o1-miniEditorial602026Source
9claude-35-sonnetEditorial59.42026Source
10grok-2Editorial562026Source
11llama-31-405bEditorial50.72026Source
12claude-3-opusEditorial50.42026Source
13gpt-4oEditorial49.92026Source
14gpt-4-turboEditorial49.32026Source
15gemini-15-proEditorial46.22026Source
16llama-31-70bEditorial41.72026Source
17gpt-4o-miniEditorial40.22026Source

Submit a Result