CMMMU (Validation): leaderboard
CMMMU (Chinese Massive Multi-discipline Multimodal Understanding) benchmark — overall accuracy on the validation split.
Metric: Validation Overall (%). Source: cmmmu-benchmark.github.io. Saturation forecast: Estimated already saturated. 35 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 52.2 | #343 |
| 2 | vicuna-13B | 35.1 | #1469 |
| 3 | GPT-4 | 32.9 | #438 |
| 4 | Llama 2 7B | 30.7 | #1520 |
| 5 | Baichuan-7B | 26 | #1527 |
| 6 | Yi 6B (Base) | 25.6 | #1472 |
| 7 | Qwen-7B | 24.7 | #1402 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cmmmu-validation · How It Works · Data refreshed daily, snapshot 2026-10-20.