Wikibench

OmniDocBench 1.5

reasoning

OmniDocBench 1.5 is a comprehensive benchmark for evaluating multimodal large language models on document understanding tasks, including OCR, document parsing, information extraction, and visual question answering across diverse document types. Lower Overall Edit Distance scores are better.

Leaderboard

Showing 15 of 15 results

  1. MiniMax M3
    91.6%
    i
  2. Qwen3.6 Plus
    91.2%
    i
  3. Qwen3.6-35B-A3B
    89.9%
    i
  4. Qwen3.5-122B-A10B
    89.8%
    i
  5. Qwen3.5-35B-A3B
    89.3%
    i
  6. GPT-5.4
    89.1%
    i
  7. Qwen3.5-27B
    88.9%
    i
  8. Kimi K2.5
    88.8%
    i
  9. GPT-5.5 Instant
    87.5%
    i
  10. GPT-5.4 mini
    87.4%
    i
  11. GPT-5.4 nano
    75.8%
    i
  12. DiffusionGemma 26B-A4B
    31.9%
    i
  13. Gemma 4 12B
    16.4%
    i
  14. Gemini 3 Flash
    12.1%
    i
  15. Gemini 3 Pro
    11.5%
    i