CMMLU: leaderboard
11,528 four-option multiple-choice questions across 67 subjects in Chinese, from MBZUAI, LibrAI and Shanghai Jiao Tong University (2023); accuracy, with a random baseline of 25%.
Metric: 5-shot Avg Accuracy (%). Source: arxiv.org. Source: github.com. Saturation forecast: Estimated already saturated. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 72B | 89.65 | #521 |
| 2 | Qwen 1.5 110B | 88.32 | #734 |
| 3 | Qwen 2.5 72B | 85.67 | #499 |
| 4 | Qwen 1.5 72B | 83.54 | #753 |
| 5 | Qwen 1.5 32B | 82.25 | #794 |
| 6 | Qwen 1.5 7B | 72.5 | #1327 |
| 7 | GPT-4 | 70.95 | #438 |
| 8 | Llama 3.1 70B Instruct | 64.38 | #561 |
| 9 | Qwen-7B | 58.66 | #1402 |
| 10 | ChatGPT | 55.51 | #900 |
| 11 | ChatGLM2 6B | 48.8 | #1498 |
| 12 | Baichuan-7B | 44.43 | #1527 |
| 13 | falcon-40B | 41.45 | #1442 |
| 14 | LLaMA-65B | 39.8 | #1303 |
| 15 | Llama 3 70B Instruct | 36.85 | #634 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cmmlu · How It Works · Data refreshed daily, snapshot 2026-10-20.