← Benchmarks
MMMU
MMMU (Massive Multi-discipline Multimodal Understanding) is a benchmark designed to evaluate multimodal models on college-level subject knowledge and deliberate reasoning. Contains 11.5K meticulously collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering across 30 subjects and 183 subfields.
id mmmu · max 1 · 66 models reported
| # | Model | Score |
|---|---|---|
| 1 | Qwen3.6 Plus Alibaba Cloud / Qwen Team | 0.86 |
| 2 | GPT-5.1 OpenAI | 0.85 |
| 3 | GPT-5.1 Instant OpenAI | 0.85 |
| 4 | GPT-5.1 Thinking OpenAI | 0.85 |
| 5 | GPT-5 OpenAI | 0.84 |
| 6 | Qwen3.5-122B-A10B Alibaba Cloud / Qwen Team · open | 0.84 |
| 7 | Seed 1.8 ByteDance | 0.83 |
| 8 | o3 OpenAI | 0.83 |
| 9 | Qwen3.6-27B Alibaba Cloud / Qwen Team · open | 0.83 |
| 10 | Qwen3.5-27B Alibaba Cloud / Qwen Team · open | 0.82 |
| 11 | Gemini 2.5 Pro Preview 06-05 Google | 0.82 |
| 12 | Qwen3.6-35B-A3B Alibaba Cloud / Qwen Team · open | 0.82 |
| 13 | o4-mini OpenAI | 0.82 |
| 14 | Qwen3.5-35B-A3B Alibaba Cloud / Qwen Team · open | 0.81 |
| 15 | Gemini 2.5 Flash Google | 0.80 |
| 16 | Seed 2.0 Mini ByteDance | 0.80 |
| 17 | Gemini 2.5 Pro Google | 0.80 |
| 18 | EXAONE 4.5 33B LG AI Research | 0.79 |
| 19 | Step3-VL-10B StepFun · open | 0.78 |
| 20 | Grok-3 xAI | 0.78 |
| 21 | o1 OpenAI | 0.78 |
| 22 | Gemini 2.0 Flash Thinking Google | 0.75 |
| 23 | GPT-4.5 OpenAI | 0.75 |
| 24 | Command A+ Cohere · open | 0.75 |
| 25 | Claude 3.7 Sonnet Anthropic | 0.75 |
| 26 | GPT-4.1 OpenAI | 0.75 |
| 27 | Claude Sonnet 4 Anthropic | 0.74 |
| 28 | Llama 4 Maverick Meta · open | 0.73 |
| 29 | Gemini 2.5 Flash-Lite Google · open | 0.73 |
| 30 | GPT-4.1 mini OpenAI | 0.73 |
| 31 | GPT-4o OpenAI | 0.72 |
| 32 | Gemini 2.0 Flash Google | 0.71 |
| 33 | QvQ-72B-Preview Alibaba Cloud / Qwen Team · open | 0.70 |
| 34 | Qwen2.5 VL 72B Instruct Alibaba Cloud / Qwen Team · open | 0.70 |
| 35 | Kimi-k1.5 Moonshot AI | 0.70 |
| 36 | Qwen2.5 VL 32B Instruct Alibaba Cloud / Qwen Team · open | 0.70 |
| 37 | Llama 4 Scout Meta · open | 0.69 |
| 38 | Claude 3.5 Sonnet Anthropic | 0.68 |
| 39 | Gemini 2.0 Flash-Lite Google | 0.68 |
| 40 | Grok-2 xAI | 0.66 |
| 41 | Gemini 1.5 Pro Google | 0.66 |
| 42 | Pixtral Large Mistral AI · open | 0.64 |
| 43 | Grok-2 mini xAI | 0.63 |
| 44 | Mistral Small 3.2 24B Instruct Mistral AI · open | 0.63 |
| 45 | Gemini 1.5 Flash Google | 0.62 |
| 46 | Nova Pro Amazon | 0.62 |
| 47 | Llama 3.2 90B Instruct Meta · open | 0.60 |
| 48 | GPT-4o mini OpenAI | 0.59 |
| 49 | Mistral Small 3.1 24B Base Mistral AI · open | 0.59 |
| 50 | Mistral Small 3.1 24B Instruct Mistral AI · open | 0.59 |
| 51 | Qwen2.5-Omni-7B Alibaba Cloud / Qwen Team · open | 0.59 |
| 52 | Qwen2.5 VL 7B Instruct Alibaba Cloud / Qwen Team · open | 0.59 |
| 53 | Nova Lite Amazon | 0.56 |
| 54 | GPT-4.1 nano OpenAI | 0.55 |
| 55 | Phi-4-multimodal-instruct Microsoft · open | 0.55 |
| 56 | Gemini 1.5 Flash 8B Google | 0.54 |
| 57 | Grok-1.5 xAI | 0.54 |
| 58 | Grok-1.5V xAI | 0.54 |
| 59 | Pixtral-12B Mistral AI · open | 0.53 |
| 60 | DeepSeek VL2 DeepSeek · open | 0.51 |
| 61 | Llama 3.2 11B Instruct Meta · open | 0.51 |
| 62 | DeepSeek VL2 Small DeepSeek · open | 0.48 |
| 63 | Gemini 1.0 Pro Google | 0.48 |
| 64 | Phi-3.5-vision-instruct Microsoft · open | 0.43 |
| 65 | DeepSeek VL2 Tiny DeepSeek · open | 0.41 |
| 66 | GPT-3.5 Turbo OpenAI | 0.00 |