Benchmark
MMMU-Pro
A more robust multi-discipline multimodal understanding benchmark that enhances MMMU through a three-step process: filtering text-only answerable…
- Modality
- multimodal
- Categories
- multimodal, reasoning, general, vision
- Openness
- unknown
- Source
- llm_stats
- Reported scores
- 68
Reported scores
llm_stats
| Model | Organization | Reported value | Reported | Evidence |
|---|---|---|---|---|
| Claude Opus 4.6 | Anthropic | 0.773 | 2026-02-05 | evidence |
| Claude Sonnet 4.6 | Anthropic | 0.756 | 2026-02-17 | evidence |
| Command A+ | Cohere | 0.63 | 2026-05-20 | evidence |
| DiffusionGemma 26B-A4B | 0.543 | 2026-06-10 | evidence | |
| Gemini 3 Flash | 0.812 | 2025-12-17 | evidence | |
| Gemini 3 Pro | 0.81 | 2025-11-18 | evidence | |
| Gemini 3.1 Flash-Lite | 0.768 | 2026-03-03 | evidence | |
| Gemini 3.1 Pro | 0.805 | 2026-02-19 | evidence | |
| Gemini 3.5 Flash | 0.836 | 2026-05-19 | evidence | |
| Gemma 4 12B | 0.691 | 2026-05-23 | evidence | |
| Gemma 4 26B-A4B | 0.738 | 2026-04-02 | evidence | |
| Gemma 4 31B | 0.769 | 2026-04-02 | evidence | |
| Gemma 4 E2B | 0.442 | 2026-04-02 | evidence | |
| Gemma 4 E4B | 0.526 | 2026-04-02 | evidence | |
| GPT-4o | OpenAI | 0.599 | 2024-08-06 | evidence |
| GPT-5 | OpenAI | 0.784 | 2025-08-07 | evidence |
| GPT-5.2 | OpenAI | 0.795 | 2025-12-11 | evidence |
| GPT-5.4 | OpenAI | 0.812 | 2026-03-05 | evidence |
| GPT-5.4 mini | OpenAI | 0.766 | 2026-03-17 | evidence |
| GPT-5.4 nano | OpenAI | 0.661 | 2026-03-17 | evidence |
| GPT-5.5 | OpenAI | 0.832 | 2026-04-23 | evidence |
| GPT-5.5 Instant | OpenAI | 0.76 | 2026-05-05 | evidence |
| GPT-5.6 Luna | OpenAI | 0.784 | 2026-07-09 | evidence |
| GPT-5.6 Sol | OpenAI | 0.83 | 2026-07-09 | evidence |
| GPT-5.6 Terra | OpenAI | 0.807 | 2026-07-09 | evidence |
| Inkling-Small | Thinking Machines Lab | 0.74 | 2026-07-30 | evidence |
| Kimi K2.5 | Moonshot AI | 0.785 | 2026-01-27 | evidence |
| Kimi K2.6 | Moonshot AI | 0.801 | 2026-04-20 | evidence |
| Kimi K3 | Moonshot AI | 0.816 | 2026-07-16 | evidence |
| LFM2.5-VL-3B | Liquid AI | 0.305 | 2026-08-12 | evidence |
| Llama 3.2 11B Instruct | Meta | 0.33 | 2024-09-25 | evidence |
| Llama 3.2 90B Instruct | Meta | 0.452 | 2024-09-25 | evidence |
| Llama 4 Maverick | Meta | 0.596 | 2025-04-05 | evidence |
| MiMo-V2.5 | Xiaomi | 0.779 | 2026-04-22 | evidence |
| MiniMax M3 | MiniMax | 0.781 | 2026-06-01 | evidence |
| Mistral Small 4 | Mistral | 0.6 | 2026-03-16 | evidence |
| Muse Glimmer-30B | Meta | 0.74 | 2026-08-10 | evidence |
| Muse Spark | Meta | 0.804 | 2026-04-08 | evidence |
| Nova 2 Lite | Amazon | 0.618 | 2025-12-02 | evidence |
| Nova 2 Omni | Amazon | 0.614 | 2025-12-02 | evidence |
| Nova 2 Pro | Amazon | 0.635 | 2025-12-02 | evidence |
| o3 | OpenAI | 0.764 | 2025-04-16 | evidence |
| Phi-4-multimodal-instruct | Microsoft | 0.385 | 2025-02-01 | evidence |
| Qwen2-VL-72B-Instruct | Qwen | 0.462 | 2024-08-29 | evidence |
| Qwen2.5 VL 32B Instruct | Qwen | 0.495 | 2025-02-28 | evidence |
| Qwen2.5 VL 72B Instruct | Qwen | 0.511 | 2025-01-26 | evidence |
| Qwen2.5 VL 7B Instruct | Qwen | 0.383 | 2025-01-26 | evidence |
| Qwen2.5-Omni-7B | Qwen | 0.366 | 2025-03-27 | evidence |
| Qwen3 VL 235B A22B Instruct | Qwen | 0.681 | 2025-09-22 | evidence |
| Qwen3 VL 235B A22B Thinking | Qwen | 0.693 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Instruct | Qwen | 0.604 | 2025-09-22 | evidence |
| Qwen3 VL 30B A3B Thinking | Qwen | 0.63 | 2025-09-22 | evidence |
| Qwen3 VL 32B Instruct | Qwen | 0.653 | 2025-09-22 | evidence |
| Qwen3 VL 32B Thinking | Qwen | 0.681 | 2025-09-22 | evidence |
| Qwen3 VL 4B Instruct | Qwen | 0.532 | 2025-09-22 | evidence |
| Qwen3 VL 4B Thinking | Qwen | 0.57 | 2025-09-22 | evidence |
| Qwen3 VL 8B Instruct | Qwen | 0.559 | 2025-09-22 | evidence |
| Qwen3 VL 8B Thinking | Qwen | 0.604 | 2025-09-22 | evidence |
| Qwen3.5-122B-A10B | Qwen | 0.769 | 2026-02-24 | evidence |
| Qwen3.5-27B | Qwen | 0.75 | 2026-02-24 | evidence |
| Qwen3.5-35B-A3B | Qwen | 0.751 | 2026-02-24 | evidence |
| Qwen3.6 Plus | Qwen | 0.788 | 2026-04-02 | evidence |
| Qwen3.6-27B | Qwen | 0.758 | 2026-04-21 | evidence |
| Qwen3.6-35B-A3B | Qwen | 0.753 | 2026-04-16 | evidence |
| Qwen3.7-Plus | Qwen | 0.79 | 2026-05-31 | evidence |
| Qwen3.8 Max | Qwen | 0.823 | 2026-08-02 | evidence |
| Seed 2.1 Pro | ByteDance | 0.827 | 2026-06-24 | evidence |
| Seed 2.1 Turbo | ByteDance | 0.822 | 2026-06-24 | evidence |
Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.