HELM · GPQA
The Frontier
Best score over time · one chart, every benchmark
Distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with HELM · GPQA
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Full rankings
34 models tested · sorted by score
| # | Model | Score |
|---|---|---|
| 1 |  Gemini 3 Pro | 80.3 |
| 2 |  GPT-5 Chat | 79.1 |
| 3 | GPT-5 Mini | 75.6 |
| 4 | o3 | 75.3 |
| 5 | Gemini 2.5 Pro | 74.9 |
| 6 | o4 Mini | 73.5 |
| 7 |  Grok 4 | 72.6 |
| 8 | gpt-oss-120b | 68.4 |
| 9 | GPT-5 Nano | 67.9 |
| 10 | Grok 3 Mini Beta | 67.5 |
| 11 |  R1 0528 | 66.6 |
| 12 |  Palmyra X5 | 66.1 |
| 13 | GPT-4.1 | 65.9 |
| 14 |  Kimi K2 0711 | 65.2 |
| 15 | Grok 3 Beta | 65.0 |
| 16 |  Qwen3 Next 80B A3B Thinking | 63.0 |
| 17 | GPT-4.1 Mini | 61.4 |
| 18 |  Claude 3.7 Sonnet | 60.8 |
| 19 | gpt-oss-20b | 59.4 |
| 20 | Claude 3.5 Sonnet | 56.5 |
| 21 | Gemini 2.0 Flash | 55.6 |
| 22 | DeepSeek V3 | 53.8 |
| 23 | Gemini 1.5 Pro (Feb 2024) | 53.4 |
| 24 | GPT-4o (2024-11-20) | 52.0 |
| 25 | GPT-4.1 Nano | 50.7 |
| 26 | Gemini 2.0 Flash Lite | 50.0 |
| 27 | GPT-5.1 | 44.2 |
| 28 | Gemini 1.5 Flash (May 2024) | 43.7 |
| 29 |  Mistral Large 2411 | 43.5 |
| 30 | Mistral Small 3.1 24B | 39.2 |
| 31 | Gemini 2.5 Flash | 39.0 |
| 32 | GPT-4o-mini (2024-07-18) | 36.8 |
| 33 | Claude 3.5 Haiku | 36.3 |
| 34 | Gemini 2.5 Flash Lite | 30.9 |
Frequently asked
Pulled from the HELM · GPQA dataset · updated daily
What does HELM · GPQA measure?
HELM · GPQA is a knowledge benchmark in the BenchGecko catalog. 34 AI models have been tested on it. Scores range from 30.9 to 80.3 out of 100.
Which model leads on HELM · GPQA?
Gemini 3 Pro from Google DeepMind leads HELM · GPQA with a score of 80.3. The median score across 34 tested models is 61.1.
Is HELM · GPQA saturated?
No · the top score is 80.3 out of 100 (80%). There is still meaningful room for improvement on HELM · GPQA.
Does HELM · GPQA predict performance on other benchmarks?
Yes · HELM · GPQA scores correlate 0.97 with Cybench across 5 shared models. Models that do well on HELM · GPQA tend to do well on Cybench.
How often is HELM · GPQA data refreshed?
BenchGecko pulls updates daily. New model scores on HELM · GPQA appear as soon as they are published by Epoch AI or the model provider.
More knowledge benchmarks
Same category · related evaluations