Wikibench

MMBench-V1.1

reasoning

Version 1.1 of MMBench, an improved bilingual benchmark for assessing multi-modal capabilities of vision-language models through multiple-choice questions in both English and Chinese, providing systematic evaluation across diverse vision-language tasks.

Leaderboard

Showing 18 of 18 results

  1. Qwen3.5-122B-A10B
    92.8%
    i
  2. Qwen3.6-35B-A3B
    92.8%
    i
  3. Qwen3.5-27B
    92.6%
    i
  4. Qwen3.6-27B
    92.3%
    i
  5. Qwen3.5-35B-A3B
    91.5%
    i
  6. Qwen3 VL 32B Thinking
    90.8%
    i
  7. Qwen3 VL 235B A22B Thinking
    90.6%
    i
  8. Qwen3 VL 235B A22B Instruct
    89.9%
    i
  9. Qwen3 VL 30B A3B Thinking
    88.9%
    i
  10. Qwen3 VL 8B Thinking
    87.5%
    i
  11. Qwen3 VL 30B A3B Instruct
    87.0%
    i
  12. Qwen3 VL 4B Thinking
    86.7%
    i
  13. Qwen3 VL 4B Instruct
    85.1%
    i
  14. Qwen3 VL 8B Instruct
    85.0%
    i
  15. Qwen2.5-Omni-7B
    81.8%
    i
  16. DeepSeek VL2 Small
    79.3%
    i
  17. DeepSeek VL2
    79.2%
    i
  18. DeepSeek VL2 Tiny
    68.3%
    i