Wikibench

MultiPL-E

language

MultiPL-E is a scalable and extensible system for translating unit test-driven code generation benchmarks to multiple programming languages. It extends HumanEval and MBPP Python benchmarks to 18 additional programming languages, enabling evaluation of neural code generation models across diverse programming paradigms and language features.

Leaderboard

Showing 13 of 13 results

  1. Qwen3-235B-A22B-Instruct-2507
    87.9%
    i
  2. Qwen3-Next-80B-A3B-Instruct
    87.8%
    i
  3. Qwen3 VL 235B A22B Instruct
    86.1%
    i
  4. Kimi K2 Instruct
    85.7%
    i
  5. Kimi K2-Instruct-0905
    85.7%
    i
  6. Qwen2.5 32B Instruct
    75.4%
    i
  7. Qwen2.5 72B Instruct
    75.1%
    i
  8. Qwen2.5 14B Instruct
    72.8%
    i
  9. Qwen2.5 7B Instruct
    70.4%
    i
  10. Qwen2 72B Instruct
    69.2%
    i
  11. Qwen3 235B A22B
    65.9%
    i
  12. Qwen2.5-Omni-7B
    65.8%
    i
  13. Qwen2 7B Instruct
    59.1%
    i