Wikibench

HumanEval

coding

A benchmark that measures functional correctness for synthesizing programs from docstrings, consisting of 164 original programming problems assessing language comprehension, algorithms, and simple mathematics

Leaderboard

Showing 20 of 72 results

  1. MiniCPM-SALA
    95.1%
    i
  2. Claude Opus 4.7
    95.0%
    i
  3. Kimi K2 0905
    94.5%
    i
  4. Claude 3.5 Sonnet
    93.7%
    i
  5. GPT-5
    93.4%
    i
  6. Kimi K2 Instruct
    93.3%
    i
  7. Qwen2.5-Coder 32B Instruct
    92.7%
    i
  8. o1
    92.4%
    i
  9. o1-mini
    92.4%
    i
  10. Sarvam-30B
    92.1%
    i
  11. Claude 3.5 Sonnet
    92.0%
    i
  12. Mistral Large 2
    92.0%
    i
  13. Claude 3.5 Sonnet
    92.0%
    i
  14. Mistral Large 2
    92.0%
    i
  15. Qwen2.5 VL 32B Instruct
    91.5%
    i
  16. GPT-4o
    90.2%
    i
  17. GPT-4o
    90.2%
    i
  18. Granite 3.3 8B Base
    89.7%
    i
  19. Granite 3.3 8B Instruct
    89.7%
    i
  20. Gemini Diffusion
    89.6%
    i