Skip to content

Toolathlon-Verified

A verified tool-use benchmark variant for completing multi-step workflows with external tools.

22 models

Higher is better.

  1. 1Claude Opus 5Anthropic80.6
  2. 2GLM-5.3-FlashZ.AI78.4
  3. 3Claude Opus 5.5Anthropic77.8
  4. 4Claude Fable 5.1Anthropic77.8
  5. 5Claude Sonnet 5.5Anthropic77.8
  6. 6MiMo-V2.6-ProXiaomi76.9
  7. 7Step 5 PreviewStepFun74.1
  8. 8DeepSeek V4 Pro 0813DeepSeek74.1
  9. 9Hy4 previewTencent74.1
  10. 10MiMo-V2.6-FlashXiaomi73.6
  11. 11Qwen3.8-Flash-NextAlibaba73.5
  12. 12Kimi K3Moonshot AI73.2
  13. 13GLM-5.3Z.AI73.0
  14. 14Muse Spark OpenMeta72.8
  15. 15Qwen3.8 MaxAlibaba72.5
  16. 16Ornith-1.5-397BOrnith AI71.2
  17. 17DeepSeek V4 Flash 0731DeepSeek70.3
  18. 18dots3-note PreviewDots Studio55.6
  19. 19Inkling-SmallThinking Machines Lab54.4
  20. 20Laguna S 2.1Poolside49.7
  21. 21Ornith-1.5-35B-A3BOrnith AI48.7
  22. 22Ornith-1.5-9BOrnith AI41.2
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
Interactive tool-use score
Tasks
Verified multi-tool workflows
Authors
Moonshot AI (2026)

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