Bowen Ye et al.

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.

In index
Agentic
Source ↗

Models39

Top result81.4%Ornith-1.5-397B

Top-3 spread4.3 pts

Year2026

Result and price

Results

39 results

#ModelResult

01OAOrnith-1.5-397B81.4%

02LAK-EXAONE 2.077.7%

03DAOrnith-1.0-397B77.1%

MiniMax M374.5%

05DSdots3-note Preview73.4%

06OAOrnith-1.5-35B-A3B72.5%

Qwen3.6-27B72.4%

Claude Opus 4.670.4%

09DAOrnith-1.0-35B69.8%

Qwen3.6-35B-A3B68.7%

Claude Sonnet 4.667.8%

12STStep 3.7 Flash67.1%

13OAOrnith-1.5-9B66.5%

Qwen3.7 Max65.2%

15INLLaDA2.2-flash64.2%

Muse Spark63.8%

MiMo-V2.5-Pro63.8%

18DAOrnith-1.0-9B63.1%

19LILFM2.5-2.6B62.9%

Qwen3.7 Plus62.7%

21MAKimi K2.662.3%

GLM-5.162.3%

MiMo-V2.562.3%

GPT-5.460.3%

Claude Opus 4.559.6%

Qwen3.6 Plus58.8%

Gemini 3.1 Pro57.8%

MiMo-V2-Pro57.8%

GLM-557.7%

30INLLaDA2.2-mini57.2%

Qwen3.5 397B56.8%

GLM-5-Turbo55.8%

GLM-5V-Turbo53.8%

34MAKimi K2.552.3%

Gemini 3 Flash49.2%

MiniMax M2.748.7%

MiMo-V2-Omni45.2%

DeepSeek V3.240.2%

Nemotron 3 Super 100B5.5%