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

Step Accuracy on AndroidControl High Level v2

64.3Pass@1

GUI-Libra-8B

Updated 7d ago

Evaluation Results

Method	Links
GUI-Libra-8B 2026.02		64.3	70.6
GUI-Libra-4B 2026.02		62.3	68.6
GPT-5 + UGround-v1-7B 2026.02		61.3	69.4
GUI-Libra-7B 2026.02		59.3	67.3
Qwen3-VL-32B 2026.02		58.8	69.6
GPT-4.1 + UGround-v1-7B 2026.02		57.5	63.3
GUI-Libra-3B 2026.02		57.3	67.1
GPT-4o + UGround-v1-7B 2026.02		57	66.3
Qwen2.5-VL-72B 2026.02		56.5	72.9
Qwen3-VL-8B 2026.02		54.8	66.1
GPT-5-mini + UGround-v1-7B 2026.02		52.8	58.8
Qwen3-VL-4B 2026.02		49.3	63.3
Qwen2.5-VL-32B 2026.02		49	66.6
Qwen2.5-VL-7B 2026.02		46.5	58.5
UI-TARS-1.5-7B 2026.02		45.2	63.1
GUI-R1-3B 2026.02		40	54
GUI-R1-7B 2026.02		39.7	56.3
Aguvis-7B 2026.02		37.7	43.7
GLM-4.1V-9B-Thinking 2026.02		37.2	49
Qwen2.5-VL-3B 2026.02		36.4	50.8