Wikibench

OCRBench-V2 (en)

vision

OCRBench v2 English subset: Enhanced benchmark for evaluating Large Multimodal Models on visual text localization and reasoning with English text content

Leaderboard

Showing 12 of 12 results

  1. Qwen3 VL 32B Thinking
    68.4%
    i
  2. Qwen3 VL 32B Instruct
    67.4%
    i
  3. Qwen3 VL 235B A22B Instruct
    67.1%
    i
  4. Qwen3 VL 235B A22B Thinking
    66.8%
    i
  5. Qwen3 VL 8B Instruct
    65.4%
    i
  6. Qwen3 VL 8B Thinking
    63.9%
    i
  7. Qwen3 VL 4B Instruct
    63.7%
    i
  8. Qwen3 VL 30B A3B Instruct
    63.2%
    i
  9. Qwen3 VL 30B A3B Thinking
    62.6%
    i
  10. Qwen3 VL 4B Thinking
    61.8%
    i
  11. Qwen2.5 VL 72B Instruct
    61.5%
    i
  12. Qwen2.5 VL 32B Instruct
    57.2%
    i