Skip to content

Toolathlon

A tool-use benchmark focused on selecting, sequencing, and completing tasks with external tools.

22 models

Higher is better.

  1. 1Muse Spark 1.1Meta75.6
  2. 2Claude Opus 4.8Anthropic59.9
  3. 3GPT-5.6 SolOpenAI58.0
  4. 4Gemini 3.5 FlashGoogle56.5
  5. 5GPT-5.5OpenAI55.6
  6. 6GPT-5.4OpenAI54.6
  7. 7GPT-5.6 LunaOpenAI53.4
  8. 8GPT-5.6 TerraOpenAI53.1
  9. 9DeepSeek V4 Pro 0813DeepSeek51.8
  10. 10Kimi K2.6Moonshot AI50.0
  11. 11Step 3.7 FlashStepFun49.5
  12. 12GLM-5.2Z.AI48.2
  13. 13DeepSeek V4 Flash 0731DeepSeek47.8
  14. 14MiniMax M2.7MiniMax46.3
  15. 15Claude Opus 4.5Anthropic43.5
  16. 16GPT-5.4 miniOpenAI42.9
  17. 17Qwen3.6 PlusAlibaba39.8
  18. 18GLM-5Z.AI38.0
  19. 19Qwen3.5 397BAlibaba36.3
  20. 20GPT-5.4 nanoOpenAI35.5
  21. 21Kimi K2.5Moonshot AI27.8
  22. 22Qwen3.6-35B-A3BAlibaba26.9
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Interactive tool-calling evaluation
Tasks
Multi-tool workflows
Authors
OpenAI (2026)

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.