Skip to content

Massive Multi-discipline Multimodal Understanding

A broad multimodal reasoning benchmark spanning charts, diagrams, tables, and academic visual question answering.

21 models

Higher is better.

  1. 1Gemini 3 ProGoogle87.2
  2. 2GPT-5.2OpenAI86.7
  3. 3Qwen3.6 PlusAlibaba86.0
  4. 4Qwen3.5 397BAlibaba85.0
  5. 5Kimi K2.5Moonshot AI84.3
  6. 6Qwen3.5-122B-A10BAlibaba83.9
  7. 7Qwen3.6-27BAlibaba82.9
  8. 8Qwen3.5-27BAlibaba82.3
  9. 9Qwen3.6-35B-A3BAlibaba81.7
  10. 10Qwen3.5-35B-A3BAlibaba81.4
  11. 11Claude Opus 4.5Anthropic80.7
  12. 12Command A+Cohere75.1
  13. 13Claude 4 OpusAnthropic73.7
  14. 14Claude 3.7 SonnetAnthropic71.8
  15. 15Nemotron 3 Nano Omni 30B A3BNVIDIA70.8
  16. 16Gemini 2.0 FlashGoogle69.3
  17. 17Grok 2xAI66.1
  18. 18Mistral Small 3.1Mistral64.0
  19. 19Gemini 1.5 FlashGoogle58.3
  20. 20LFM2.5-VL-3BLiquid AI48.4
  21. 21LFM2.5-VL-450MLiquid AI32.7
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Image + text question answering
Tasks
Multimodal academic reasoning
Authors
MMMU authors (2024)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.