Wikibench

Humanity's Last Exam

math

Humanity's Last Exam (HLE) is a multi-modal academic benchmark with 2,500 questions across mathematics, humanities, and natural sciences, designed to test LLM capabilities at the frontier of human knowledge with unambiguous, verifiable solutions

Leaderboard

Showing 20 of 82 results

  1. Claude Mythos Preview
    64.7%
    i
  2. Claude Fable 5
    64.5%
    i
  3. Muse Spark
    58.4%
    i
  4. Claude Opus 4.8
    57.9%
    i
  5. GPT-5.5 Pro
    57.2%
    i
  6. Claude Opus 4.7
    54.7%
    i
  7. Claude Opus 4.6
    53.1%
    i
  8. GLM-5.1
    52.3%
    i
  9. GPT-5.5
    52.2%
    i
  10. Gemini 3.1 Pro
    51.4%
    i
  11. Kimi K2-Thinking-0905
    51.0%
    i
  12. Grok-4 Heavy
    50.7%
    i
  13. Kimi K2.5
    50.2%
    i
  14. Claude Sonnet 4.6
    49.0%
    i
  15. Qwen3.5-27B
    48.5%
    i
  16. DeepSeek-V4-Pro-Max
    48.2%
    i
  17. Qwen3.5-122B-A10B
    47.5%
    i
  18. Qwen3.5-35B-A3B
    47.4%
    i
  19. Gemini 3 Pro
    45.8%
    i
  20. DeepSeek-V4-Flash-Max
    45.1%
    i