List your company

Fine-Grained Action Execution on OSExpert-Eval

28GIMP Success RateOSExpert (Explore w/ Qwen-3-VL-8B)
-1.126.441421.56Mar 9, 2026

Evaluation results

MethodLinks
OSExpert (Explore w/ Qwen-3-VL-8B)Agent Type=Agentic Framework2026.032826
OSExpert (Explore w/ GPT-5 & UI-TARS)Agent Type=Agentic Framework2026.032826
OSExpert (Explore w/ Qwen-3-VL-8B, wo/ Fine Skills)Agent Type=Agentic Framework2026.031410
OpenCUA-7BAgent Type=Specialized Model2026.0375
Qwen-3-VL-8BAgent Type=General Model2026.03710
Computer-Use-PreviewAgent Type=Specialized Model2026.0300
CoAct-1Agent Type=Agentic Framework2026.0305
Agent-S3 w/ GPT-5Agent Type=Agentic Framework2026.03010