List your company

GUI Navigation and Action on OS World (test)

2,056Success Rate (Avg)ANCHOR
-79.12475.191,029.51,583.81Oct 30, 2024Jan 30, 2025May 3, 2025Aug 3, 2025Nov 4, 2025Feb 4, 2026May 8, 2026

Evaluation results

MethodLinks
ANCHORModel=Qwen3-VL-8B, Strategies=ANCHOR2026.022,05636.3713.7918.5238.4657.14—26.6718.1837.5—7.1450
Qwen3-VL-8BStrategies=Task-Driven2026.021,77545.4510.3418.5238.4642.85—23.3327.2725—5.7125
Qwen3-VL-8BStrategies=Zero-Shot2026.021,68227.276.914.8130.7742.85—23.3327.2737.5—7.1425
Qwen3-VL-8BStrategies=Human Data2026.021,63536.3713.7914.8123.0742.85—13.3336.3637.5—5.7125
ANCHORModel=Qwen2.5-VL-7B, Strategies=ANCHOR2026.027949.093.453.715.3814.29—1027.2725—4.290
ANCHORModel=GLM4.1V-9B, Strategies=ANCHOR2026.0270103.457.4115.3814.29—109.0912.5—2.8512.5
Qwen2.5-VL-7BStrategies=Task-Driven2026.025619.0903.715.380—1018.1812.5—2.860
GLM4.1V-9BStrategies=Task-Driven2026.025149.093.453.77.690—6.6718.180—2.8612.5
GLM4.1V-9BStrategies=Human Data2026.025149.0903.77.690—6.6718.1812.5—4.290
Qwen2.5-VL-7BStrategies=Human Data2026.024679.0907.417.6914.29—6.6700—4.290
Qwen2.5-VL-7BStrategies=Zero-Shot2026.029300000—3.3300—1.430
Agent Alphabase model=GPT-5.2, action chunking=true2026.0277.2995.8387.2378.7291.370.58—67.3910096.15—59.1480
Agent s3strategy=bBoN, N=102026.0272.5879.1787.2372.2686.8362.94—62.9682.6173.08—63.973.33
Human2024.1072.367561.780.8573.9170.5946.6778.2673.9173.0873.27——
UiPath Screen Agent2026.0267.1470.8380.4370.1969.5760.82—69.4878.2669.23—52.9773.33
OS-Symphony2026.0265.7778.2676.648.8178.1361.53—62.9678.2680.77—55.380
GBOX Agent2026.0264.2270.8374.4763.7468.1846.29—62.9691.376.92—49.7366.67
GTA12026.0263.4179.1763.8365.5360.7457.88—58.6182.6176.92—50.9180
claude-sonnet-4-52026.0262.8870.8372.346882.4858.18—62.9673.9153.85—49.5460
CoACT-12026.0260.767570.2150.2873.9171.94—54.2678.2665.38—47.8773.33
GLM4.1V-9BStrategies=Zero-Shot2026.024700000—3.3300—00
Claude-sonnet-4-5Type=General, Open source=False2026.0542.9————————————
doubao-1-5-thinking-vision-proType=Specialized, Open source=False2026.0539————————————
UI-TARS-1.5-7BType=Specialized, Open source=True2026.0524.5————————————
OpenCUA-7BType=Specialized, Open source=True2026.0524.3————————————
UI-TARS-72B-DPOType=Specialized, Open source=True2026.0524————————————
Lite-GUI-30B-A3BType=Specialized, Open source=True2026.0522.7————————————
Opencua-Qwen2-7BType=Specialized, Open source=True2026.0519.9————————————
OpenCUA-A3BType=Specialized, Open source=True2026.0516.9————————————
Qwen3-VL-30B-A3B-InstructType=General, Open source=True2026.0516.7————————————
GPT-4o + OS-Atlas-Base-7BGrounding Module=OS-Atlas-Base-7B, Backbone Model=Qwen2-VL-7B2024.1014.63254.2617.028.729.4126.6719.5717.3919.238.91——
Lite-GUI-2BType=Specialized, Open source=True2026.0513.24————————————
GPT-4o + OS-Atlas-Base-4BGrounding Module=OS-Atlas-Base-4B, Backbone Model=InternVL-2-4B2024.1011.6520.832.2314.898.723.5213.3315.2213.0415.387.92——
Kimi-VL-A3BType=General, Open source=True2026.059.7————————————
GPT-4o + SeeClickGrounding Module=SeeClick2024.109.2116.67012.764.3523.526.6710.868.711.547.92——
GPT-o3Type=General, Open source=False2026.059.1————————————
Qwen3-VL-2B-InstructType=General, Open source=True2026.056.04————————————
GPT-4oGrounding Module=Native/Internal2024.105.038.3306.774.3516.104.354.353.855.58——
GPT-4o + SoMGrounding Module=Set-of-Mark (SoM)2024.104.5920.8306.774.356.5204.354.3503.6——
Qwen2.5-VL-72B-instructType=General, Open source=True2026.054.4————————————
Qwen2.5-VL-32B-instructType=General, Open source=True2026.053————————————