Share your thoughts, 1 month free Claude Pro on usSee more

GUI Agent Success on WebVoyager, Online-Mind2Web, and DeepShop (test average)

69.3Average Success Rate

Gemini computer-use-preview

Updated 19d ago

Evaluation Results

Method	Links
Gemini computer-use-preview 2026.05		69.3
Orchard-GUI-4B (SFT + RL) 2026.05		68.4
GPT-5 (SoM) 2026.05		65.8
Gemini-3-flash (Axtree) 2026.05		61.9
o3 (SoM) 2026.05		61.5
Qwen3-VL-235B-A22B-Thinking 2026.05		61.2
Orchard-GUI-4B-SFT 2026.05		52
MolmoWeb-8B 2026.05		51.9
Gemini-3-flash (Axtree) 2026.05		51.4
OpenAI computer-use-preview 2026.05		51.3
GPT-5 (Axtree) 2026.05		51.1
MolmoWeb-4B 2026.05		47.4
Fara-7B 2026.05		44.6
GLM-4.1V-9B-Thinking 2026.05		44.2
GPT-4o (SoM) 2026.05		38.6
Qwen3-VL-4B-Thinking 2026.05		38.1
UI-TARS-1.5-7B 2026.05		36.4