Skip to content

MMLU-ProX

A multilingual extension of professional-level academic evaluation across many languages.

16 models

Higher is better.

  1. 1Qwen3.7 MaxAlibaba87.0
  2. 2Claude Opus 4.5Anthropic85.7
  3. 3Qwen3.7 PlusAlibaba85.4
  4. 4Qwen3.5 397BAlibaba84.7
  5. 5Qwen3.6 PlusAlibaba84.7
  6. 6GLM-5Z.AI83.1
  7. 7Nemotron 3 UltraNVIDIA83.0
  8. 8Kimi K2.5Moonshot AI82.3
  9. 9Qwen3.5-27BAlibaba82.2
  10. 10Qwen3.5-122B-A10BAlibaba82.2
  11. 11Qwen3.5-35B-A3BAlibaba81.0
  12. 12Qwen3 235B 2507Alibaba79.4
  13. 13GPT-4.1OpenAI72.7
  14. 14DeepSeek V3 0324DeepSeek70.5
  15. 15GPT-4oOpenAI61.1
  16. 16Phi-4Microsoft49.9
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Multilingual multiple choice
Tasks
Multilingual professional QA
Authors
MMLU-ProX authors (2025)

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