multimodal Multimodal · llm-stats.com
MMMU-Pro
Harder multimodal reasoning successor of MMMU.
18
Getestete Modelle
GPT-6 Astra
Top-Modell
86.4%
Top-Score
OpenAI
Anbieter
| # | Modell | MMMU-Pro |
|---|---|---|
| 1 | GPT-6 Astra OpenAI | 86.4% |
| 2 | GPT-5.5 OpenAI | 83.2% |
| 3 | GPT-5.6 Sol OpenAI | 83% |
| 4 | Qwen3.8 Max Alibaba | 82.3% |
| 5 | Kimi K3 Moonshot AI | 81.6% |
| 6 | GPT-5.4 OpenAI | 81.2% |
| 7 | Gemini 3 Flash Google | 81.2% |
| 8 | Gemini 3 Pro Google | 81% |
| 9 | GPT-5.6 Terra OpenAI | 80.7% |
| 10 | Gemini 3.1 Pro Google | 80.5% |
| 11 | Muse Spark Meta | 80.4% |
| 12 | Kimi K2.6 Moonshot AI | 80.1% |
| 13 | GPT-5.2 OpenAI | 79.5% |
| 14 | Qwen3.7 Plus Alibaba | 79% |
| 15 | Qwen3.6 Plus Alibaba | 78.8% |
| 16 | GPT-5.6 Luna OpenAI | 78.4% |
| 17 | Claude Opus 4.6 Anthropic | 77.3% |
| 18 | Claude Sonnet 4.6 Anthropic | 75.6% |
Benchmark-Daten aggregiert aus llm-stats.com. Siehe Methodik.