Wikibench

Toolathlon

reasoning

Tool Decathlon is a comprehensive benchmark for evaluating AI agents' ability to use multiple tools across diverse task categories. It measures proficiency in tool selection, sequencing, and execution across ten different tool-use scenarios.

Leaderboard

Showing 20 of 20 results

  1. Claude Opus 4.8
    59.9%
    i
  2. Gemini 3.5 Flash
    56.5%
    i
  3. GPT-5.5
    55.6%
    i
  4. GPT-5.4
    54.6%
    i
  5. DeepSeek-V4-Pro-Max
    51.8%
    i
  6. Kimi K2.6
    50.0%
    i
  7. Gemini 3 Flash
    49.4%
    i
  8. DeepSeek-V4-Flash-Max
    47.8%
    i
  9. GPT-5.2
    46.3%
    i
  10. MiniMax M2.7
    46.3%
    i
  11. MiniMax M2.1
    43.5%
    i
  12. GPT-5.4 mini
    42.9%
    i
  13. GLM-5.1
    40.7%
    i
  14. Qwen3.6 Plus
    39.8%
    i
  15. Qwen3.5-397B-A17B
    38.3%
    i
  16. GPT-5.4 nano
    35.5%
    i
  17. DeepSeek-V3.2 (Thinking)
    35.2%
    i
  18. DeepSeek-V3.2
    35.2%
    i
  19. DeepSeek-V3.2-Speciale
    35.2%
    i
  20. Qwen3.6-35B-A3B
    26.9%
    i