GPQA
The Frontier
Best score over time · one chart, every benchmark
Full rankings
72 models tested · sorted by score
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with GPQA
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About GPQA
What does GPQA measure?
GPQA is a knowledge benchmark in the BenchGecko catalog. 72 AI models have been tested on it. Scores range from 0.5 to 19.7 out of 100.
Which model leads on GPQA?
Meta Llama 3 8B from Meta leads GPQA with a score of 19.7. The median score across 72 tested models is 4.8.
Is GPQA saturated?
No · the top score is 19.7 out of 100 (20%). There is still meaningful room for improvement on GPQA.
Does GPQA predict performance on other benchmarks?
Yes · GPQA scores correlate 0.97 with CMMLU across 5 shared models. Models that do well on GPQA tend to do well on CMMLU.
How often is GPQA data refreshed?
BenchGecko pulls updates daily. New model scores on GPQA appear as soon as they are published by Epoch AI or the model provider.
- Category
- Knowledge
- Max score
- 100
- Models
- 72
- Updated
- 2025-10-17
Top on GPQA
Meta Llama 3 8B · 19.7Qwen2.5 72B Instruct Abliterated · 19.4Qwen2-72B · 19.2DeepSeek R1 Distill Qwen 14B · 18.3WizardLM-2 8x22B · 17.6More knowledge benchmarks
Same category · related evaluations