Wikibench

PolyMATH

math

Polymath is a challenging multi-modal mathematical reasoning benchmark designed to evaluate the general cognitive reasoning abilities of Multi-modal Large Language Models (MLLMs). The benchmark comprises 5,000 manually collected high-quality images of cognitive textual and visual challenges across 10 distinct categories, including pattern recognition, spatial reasoning, and relative reasoning.

Leaderboard

Showing 20 of 22 results

  1. Qwen3.7 Max
    86.5%
    i
  2. Qwen3.6 Plus
    77.4%
    i
  3. Qwen3.5-397B-A17B
    73.3%
    i
  4. Qwen3.5-27B
    71.2%
    i
  5. Qwen3.5-122B-A10B
    68.9%
    i
  6. Qwen3.5-35B-A3B
    64.4%
    i
  7. Qwen3-235B-A22B-Thinking-2507
    60.1%
    i
  8. Qwen3.5-9B
    57.3%
    i
  9. Qwen3-Next-80B-A3B-Thinking
    56.3%
    i
  10. Qwen3 VL 32B Thinking
    52.0%
    i
  11. Qwen3 VL 30B A3B Thinking
    51.7%
    i
  12. Qwen3.5-4B
    51.1%
    i
  13. Qwen3-235B-A22B-Instruct-2507
    50.2%
    i
  14. Qwen3 VL 8B Thinking
    47.5%
    i
  15. Qwen3-Next-80B-A3B-Instruct
    45.9%
    i
  16. Qwen3 VL 4B Thinking
    44.6%
    i
  17. Qwen3 VL 30B A3B Instruct
    44.3%
    i
  18. Qwen3 VL 32B Instruct
    40.5%
    i
  19. Qwen3 VL 8B Instruct
    30.4%
    i
  20. Qwen3 VL 4B Instruct
    28.8%
    i