HELM Capabilities - MMLU-Pro — leaderboard
Metric: COT correct. Source: crfm.stanford.edu. 51 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 03-25) | 86.3 |
| 2 | O3 (2025-04-16) | 85.9 |
| 3 | Claude Opus 4 (20250514) | 85.9 |
| 4 | Grok 4 | 85.1 |
| 5 | Claude Sonnet 4 (20250514) | 84.3 |
| 6 | O4 Mini (2025-04-16) | 82 |
| 7 | Kimi K2 | 81.9 |
| 8 | Qwen3 235B A22B FP8 Throughput | 81.7 |
| 9 | GPT-4.1 (2025-04-14) | 81.1 |
| 10 | Llama 4 Maverick Instruct FP8 | 81 |
| 11 | Palmyra X5 | 80.4 |
| 12 | Grok 3 Mini Beta | 79.9 |
| 13 | DeepSeek R1 0528 | 79.3 |
| 14 | Grok 3 | 78.8 |
| 15 | Claude 3.7 Sonnet (20250219) | 78.4 |
Interactive version: theaggregate.ai/benchmark?slug=helm-capabilities-mmlu-pro · How the rankings work · Data refreshed daily, snapshot 2026-07-22.