OSWorld
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
OSWorld-G
72.7Average Score
164
OSWorld-G (test)
78.4Element Accuracy
52
OSWorld
75OS Success Rate
45
OSWorld 361 tasks (Verified)
79.17OS Success Rate
43
OSWorld
64.29Success Rate
42
OSWorld 1.0 (test)
91.67Success Rate (OS)
42
OSWorld Verified (test)
61.92Overall Success Rate
40
OSWorld G-Refine v1.0 (test)
76.1Overall Success Rate
37
OSWorld
40.3SR (15 Steps)
34
OSWorld Verified
91.7OS Success Rate
32
OSWorld
86.8Overall Success Rate
28
OSWorld
42.5Average Accuracy
27
OSWorld-G-Refine
76.4Accuracy
24
OSWorld (verified)
64.8Office Task Score
24
OSWorld
50.11Overall Success Rate
23
OSWorld-G v1.0 (test)
67.6Accuracy
22
OSWorld
64.7Overall Score
22
OSworld G-R
76.4Accuracy
22
OSWorld-G-Refine (test)
76.4Overall Success Rate
17
OSWorld Verified
70.8OS Performance
16
OSWorld
42.9Success Rate (Max Steps: 15)
16
OSWorld-Verified (test)
61.4Pass@1
16
OSWorld
24.4Success Rate
16
OSWorld
41.49Task Accuracy
15
OSWorld
41.49Multimodal Task Accuracy
15