List your company

OS agent GUI interaction on VeriOS-Bench

47.62MC ScoreVeriOS-Agent
-1.904810.952623.8136.6674Sep 9, 2025

Evaluation results

MethodLinks
VeriOS-AgentModel Size=32B, Training Protocol=RL2025.0947.6273.2184.0994.1270.1273.83
VeriOS-AgentModel Size=7B, Training Protocol=RL2025.0946.4371.4379.5581.4848.1765.41
VeriOS-AgentModel Size=7B, Training Protocol=SFT2025.0945.2471.4377.2781.4847.5657.22
VeriOS-AgentModel Size=32B, Training Protocol=SFT2025.0945.2471.4381.8292.9569.5168.98
UI-TARS-1.5Model Size=7B2025.0923.8114.2959.0962.9658.5448.13
GLM-4.5V2025.0923.8121.435077.7863.4151.87
UI-TARSModel Size=7B, Training Protocol=SFT2025.0921.4321.4359.0970.3759.7649.73
OS-Atlas-ProModel Size=7B2025.0919.057.1422.7359.2664.6344.39
UI-TARSModel Size=7B, Training Protocol=DPO2025.0919.0514.2954.5570.3763.4149.73
UI-TARSModel Size=2B, Training Protocol=SFT2025.0916.6714.2945.4566.6741.4637.97
Qwen2.5-VLModel Size=72B2025.0916.6721.435077.7871.9554.01
GUI-OwlModel Size=32B2025.0914.2914.2954.5570.3769.5151.34
GPT-4o2025.0914.297.1431.8281.4848.7840.64
Qwen-VL-max2025.0911.914.294.5518.5226.8318.72
Qwen2.5-VLModel Size=32B2025.099.5214.2936.3677.7860.9845.45
Qwen2.5-VLModel Size=7B2025.092.380044.4437.823.53
Qwen2.5-VLModel Size=3B2025.090022.737.413.665.35