List your company

Grounding on OSWorld

64.7Overall ScoreMGA w/ GPT-5
15.19628.04840.953.752Oct 28, 2025

Evaluation results

MethodLinks
MGA w/ GPT-5Steps=50, Planner Agent=GPT-5, Memory Agent=qwen3-8b2025.1064.764.764.485.487.547.7
OS-SYMPHONY w/ GPT-5Steps=502025.1063.664.961.269.27554.9
GTA1-7b w/ GPT-5Steps=502025.1061—————
CoAct-1 w/ GPT-5Steps=1502025.1060.862.961.771.47547.8
CoAct-1 w/ GPT-5Steps=1002025.1059.962.957.971.47547.8
Claude-Sonnet-4.5Steps=502025.1058.162.457.663.270.847
OS-SYMPHONY w/ GPT5-MiniSteps=502025.1058.158.261.47573.747.4
UiPath w/ GPT-5Steps=502025.1053.749.562.171.473.937.3
UI-TARS-2Steps=1002025.1053.161.162.161.241.634.1
GTA1-7b w/ o3Steps=1002025.1053.154.660.361.262.538.3
MGA w/ o3Steps=50, Planner Agent=o3, Memory Agent=o32025.1052.94962.870.668.236.4
Jedi-7B w/ o3Steps=502025.1050.64762.169.454.234.9
GTA1-7b w/ o3Steps=502025.1048.646.144.977.658.337.1
Agent S3 w/ GPT-5-MiniSteps=502025.1047.554.646.644.962.537
OS-SYMPHONY w/ Qwen3-VL-32BSteps=502025.1046.940.953.575.158.331.2
OpenCUA-72BSteps=1002025.1044.944.749.972.561.122.1
UI-TARSSteps=1002025.1041.850.455.65141.614.6
Qwen3-VL-32B-ThinkingSteps=502025.1041—————
DeepMiner-Mano-7BSteps=1002025.1040.139.244.873.45017.2
OpenCUA-32bSteps=502025.1034.129.940.965.360.914.6
Qwen3-VL-32B-InstructSteps=502025.1032.4—————
O3Steps=502025.1017.19.420.530.637.511.8