Wikibench

MMLU-Pro

math

A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks. Features over 12,000 curated questions across 14 domains and causes a 16-33% accuracy drop compared to original MMLU.

Leaderboard

Showing 20 of 127 results

  1. Qwen3.7 Max
    89.6%
    i
  2. Qwen3.6 Plus
    88.5%
    i
  3. MiniMax M2.1
    88.0%
    i
  4. Qwen3.5-397B-A17B
    87.8%
    i
  5. DeepSeek-V4-Pro-Max
    87.5%
    i
  6. Kimi K2.5
    87.1%
    i
  7. ERNIE 5.0
    87.0%
    i
  8. Qwen3.5-122B-A10B
    86.7%
    i
  9. DeepSeek-V4-Flash-Max
    86.2%
    i
  10. Qwen3.6-27B
    86.2%
    i
  11. Qwen3.5-27B
    86.1%
    i
  12. Qwen3.5-35B-A3B
    85.3%
    i
  13. Gemma 4 31B
    85.2%
    i
  14. Qwen3.6-35B-A3B
    85.2%
    i
  15. DeepSeek-R1-0528
    85.0%
    i
  16. DeepSeek-V3.2-Exp
    85.0%
    i
  17. DeepSeek-V3.2 (Thinking)
    85.0%
    i
  18. DeepSeek-V3.2
    85.0%
    i
  19. MAI-Thinking-1
    85.0%
    i
  20. MiMo-V2-Flash
    84.9%
    i