List your company

Action Prediction on AITZ (test)

76.12Scroll AccuracyOS-Atlas
8.540826.085443.6361.1746Feb 27, 2026

Evaluation results

MethodLinks
OS-Atlas#Params=7B, Finetuned with hybrid-capabilities reasoning=false2026.0276.1275.8254.8389.888.668.6781.7577.8365.11
Auto-GUI#Params=700M, Source=Reported from Zhang et al., 2024, Finetuned with hybrid-capabilities reasoning=false2026.0261.474.5632.287.881.457.774.482.9847.69
UGround#Params=7B, Finetuned with hybrid-capabilities reasoning=false2026.0258.2280.9458.4882.5673.8558.2268.7874.5460.19
UITars#Params=7B, Finetuned with hybrid-capabilities reasoning=false2026.0256.584.8763.8785.9785.7758.2271.2978.0765.41
AriaUI#Params=3.9B, Finetuned with hybrid-capabilities reasoning=false2026.0253.7385.5160.284.280.863.776.3878.5363.56
CoME#Params=5B, Finetuned with hybrid-capabilities reasoning=false2026.0252.0783.8365.2288.283.859.5383.3378.666.98
Qwen2VL-MoE#Params=3B, Finetuned with hybrid-capabilities reasoning=true2026.0248.7580.0359.4377.87457.770.8372.9460.69
UI-S1#Params=7B, Finetuned with hybrid-capabilities reasoning=false2026.0245.678.7556.7391.287.251.1739.2869.8556.22
Qwen2VL#Params=7B, Finetuned with hybrid-capabilities reasoning=true2026.0243.2883.9755.351.851.456.9264.8771.5954.46
UITars#Params=2B, Finetuned with hybrid-capabilities reasoning=false2026.0230.1783.8155.2583.9784.7753.7861.3974.2355.63
GUI-R1#Params=7B, Finetuned with hybrid-capabilities reasoning=false2026.0229.0282.2253.8829.62745.1742.2662.8745.91
MolmoE#Params=1B, Finetuned with hybrid-capabilities reasoning=true2026.0228.1979.8433.1765.26237.0754.9665.9638.23
GUI-R1#Params=3B, Finetuned with hybrid-capabilities reasoning=false2026.0225.537950.4721.820.832.8928.7757.3240.41
ShowUI#Params=2B, Finetuned with hybrid-capabilities reasoning=false2026.0223.7953.2940.3489.685.25987.5564.2249.55
Qwen2VL#Params=2B, Finetuned with hybrid-capabilities reasoning=true2026.0223.0578.3843.7460.459.452.2247.0262.2543.8
DeepSeekVL2#Params=4.5B, Finetuned with hybrid-capabilities reasoning=true2026.0217.9475.3519.9850.946.6914.3624.2555.3622.55
SeeClick#Params=7B, Finetuned with hybrid-capabilities reasoning=false2026.0211.1469.9252.9653.85367.8855.3662.9349.11