OpenEval - GPQA CoT — leaderboard
Metric: CoT Correctness (%). Source: huggingface.co. 43 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 45.07 |
| 2 | Gemma 3 27B (IT) | 40.36 |
| 3 | Qwen 3 4B Instruct | 39.69 |
| 4 | Qwen 2.5 32B Instruct | 38.79 |
| 5 | Qwen 2 72B Instruct | 37.89 |
| 6 | Gemma 3 12B (IT) | 36.55 |
| 7 | Qwen 2.5 14B Instruct | 36.1 |
| 8 | Qwen 2.5 3B Instruct | 32.74 |
| 9 | Qwen 1.5 32B Chat | 31.39 |
| 10 | Gemma 2 27B (IT) | 30.72 |
| 11 | Qwen 1.5 14B Chat | 30.72 |
| 12 | Gemma 1.1 7B (IT) | 28.7 |
| 13 | Qwen 2.5 7B Instruct | 28.48 |
| 14 | Qwen 2 7B Instruct | 27.8 |
| 15 | gemma-7B (IT) | 27.58 |
Interactive version: theaggregate.ai/benchmark?slug=openeval-gpqa-cot · How the rankings work · Data refreshed daily, snapshot 2026-07-22.