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

Mobile Agent Evaluation on AndroidControl Low (test)

93.7Task Success Rate

Qwen2.5-VL-72B

Updated 1mo ago

Evaluation Results

Method	Links
Qwen2.5-VL-72B 2025.08		93.7	-	-
Qwen2.5-VL-32B 2025.08		93.3	-	-
CRAFT-GUI-32B 2025.08		92.7	93.6	98.9
CRAFT-GUI-32B 2025.08		92.2	93.1	98.8
UI-TARS-72B 2025.08		91.3	89.9	98.1
CRAFT-GUI-32B 2025.08		91.3	92.4	98
OS-Atlas-7B 2025.08		85.2	88	93.6
Aguvis-72B 2025.08		84.4	-	-
Qwen2-VL-7B 2025.08		82.6	86.5	91.9
InternVL-2-4B 2025.08		80.1	84.1	90.9
SeeClick 2025.08		75	73.4	93
Aria-UI 2025.08		67.3	87.7	-
Claude 2025.08		19.4	0	74.3
GPT-4o 2025.08		19.4	0	74.3
Aria-UITH 2024.12		0.6733	0.8769	-
Aria-UIIH 2024.12		0.6726	0.872	-
Aria-UI 2024.12		0.663	0.8571	-
Aria-UI 2024.12		0.5439	0.797	-
UGround 2024.12		0.4685	0.7428	-
Qwen2-VL 2024.12		0.3253	0.6424	-
SeeClick 2024.12		0.1772	0.4555	-
GPT-4o 2024.12		0.0512	0.1636	-