👁️ Multimodal · llm-stats.com
MMMU-Pro
Harder multimodal reasoning successor of MMMU.
17
测试模型数
GPT-5.5
最佳模型
83.2%
最高分
OpenAI
提供商
| # | 模型名称 | MMMU-Pro |
|---|---|---|
| 1 | GPT-5.5 OpenAI | 83.2% |
| 2 | GPT-5.6 Sol OpenAI | 83% |
| 3 | Qwen3.8 Max Alibaba | 82.3% |
| 4 | Kimi K3 Moonshot AI | 81.6% |
| 5 | GPT-5.4 OpenAI | 81.2% |
| 6 | Gemini 3 Flash Google | 81.2% |
| 7 | Gemini 3 Pro Google | 81% |
| 8 | GPT-5.6 Terra OpenAI | 80.7% |
| 9 | Gemini 3.1 Pro Google | 80.5% |
| 10 | Muse Spark Meta | 80.4% |
| 11 | Kimi K2.6 Moonshot AI | 80.1% |
| 12 | GPT-5.2 OpenAI | 79.5% |
| 13 | Qwen3.7 Plus Alibaba | 79% |
| 14 | Qwen3.6 Plus Alibaba | 78.8% |
| 15 | GPT-5.6 Luna OpenAI | 78.4% |
| 16 | Claude Opus 4.6 Anthropic | 77.3% |
| 17 | Claude Sonnet 4.6 Anthropic | 75.6% |
基准数据聚合自 llm-stats.com。详见方法论。