GTA1: GUI Test-time Scaling Agent

Benchmark Model Rank Results
computer-use-agents-on-osworld-50-stepsGTA1-7B-2507 + o3 planner/judge#3Success Rate: 43.4
computer-use-agents-on-screenspot-v2GTA1-32B#8Accuracy: 95.2
computer-use-agents-on-osworld-verifiedGTA1-32B + GPT-5 planner/judge#9Success Rate: 63.4
computer-use-agents-on-screenspot-proGTA1-32B#9Accuracy: 63.6