Skip to content

Claw-Eval

A transparent real-world autonomous-agent benchmark with 300 human-verified tasks, 2,159 rubric items, and Pass^3 scoring across general, multi-turn, and native multimodal agent tasks.

36 models

Higher is better.

  1. 1Ornith-1.5-397BOrnith AI81.4
  2. 2K-EXAONE 2.0LG AI Research77.7
  3. 3Ornith-1.0-397BDeepReinforce AI77.1
  4. 4MiniMax M3MiniMax74.5
  5. 5dots3-note PreviewDots Studio73.4
  6. 6Ornith-1.5-35B-A3BOrnith AI72.5
  7. 7Qwen3.6-27BAlibaba72.4
  8. 8Claude Opus 4.6Anthropic70.4
  9. 9Qwen3.6-35B-A3BAlibaba68.7
  10. 10Claude Sonnet 4.6Anthropic67.8
  11. 11Step 3.7 FlashStepFun67.1
  12. 12Ornith-1.5-9BOrnith AI66.5
  13. 13Qwen3.7 MaxAlibaba65.2
  14. 14LLaDA2.2-flashInclusionAI64.2
  15. 15Muse SparkMeta63.8
  16. 16MiMo-V2.5-ProXiaomi63.8
  17. 17LFM2.5-2.6BLiquid AI62.9
  18. 18Qwen3.7 PlusAlibaba62.7
  19. 19Kimi K2.6Moonshot AI62.3
  20. 20GLM-5.1Z.AI62.3
  21. 21MiMo-V2.5Xiaomi62.3
  22. 22GPT-5.4OpenAI60.3
  23. 23Claude Opus 4.5Anthropic59.6
  24. 24Qwen3.6 PlusAlibaba58.8
  25. 25Gemini 3.1 ProGoogle57.8
  26. 26MiMo-V2-ProXiaomi57.8
  27. 27GLM-5Z.AI57.7
  28. 28Qwen3.5 397BAlibaba56.8
  29. 29GLM-5-TurboZ.AI55.8
  30. 30GLM-5V-TurboZ.AI53.8
  31. 31Kimi K2.5Moonshot AI52.3
  32. 32Gemini 3 FlashGoogle49.2
  33. 33MiniMax M2.7MiniMax48.7
  34. 34MiMo-V2-OmniXiaomi45.2
  35. 35DeepSeek V3.2DeepSeek40.2
  36. 36Nemotron 3 Super 100BNVIDIA5.5
Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric
End-to-end autonomous-agent evaluation with Pass^3 scoring
Tasks
300 tasks, 2,159 rubrics
Authors
Bowen Ye, Rang Li, Qibin Yang, Yuanxin Liu, Linli Yao, Hanglong Lv, Zhihui Xie, Chenxin An, Lei Li, Lingpeng Kong, Qi Liu, Zhifang Sui, Tong Yang (2026)

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