HELM Capabilities - GPQA — leaderboard
Metric: COT correct. Source: crfm.stanford.edu. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 (2025-04-16) | 75.34 |
| 2 | Gemini 2.5 Pro (Preview 03-25) | 74.89 |
| 3 | O4 Mini (2025-04-16) | 73.54 |
| 4 | Grok 4 | 72.65 |
| 5 | Grok 3 Mini Beta | 67.49 |
| 6 | DeepSeek R1 0528 | 66.59 |
| 7 | Claude Opus 4 (20250514) | 66.59 |
| 8 | Palmyra X5 | 66.14 |
| 9 | GPT-4.1 (2025-04-14) | 65.92 |
| 10 | Kimi K2 | 65.25 |
| 11 | Grok 3 | 65.02 |
| 12 | Llama 4 Maverick Instruct FP8 | 65.02 |
| 13 | Claude Sonnet 4 (20250514) | 64.35 |
| 14 | Qwen3 235B A22B FP8 Throughput | 62.33 |
| 15 | GPT-4.1 Mini | 61.43 |
Interactive version: theaggregate.ai/benchmark?slug=helm-capabilities-gpqa · How the rankings work · Data refreshed daily, snapshot 2026-07-22.