ResearchBenchmarksGUI Agent Task Success on OSWorld Chrome (test)Follow70.6Success RateGPT-4o46.1652.50558.8565.195Jun 2, 2026Evaluation ResultsMethodMethodLinksSuccess RateGPT-4oModel Setup=+TutorialModel Setup=+Tutorial2026.0670.6o1-miniModel Setup=+TutorialModel Setup=+Tutorial2026.0658.8GPT-4oModel Setup=baselineModel Setup=baseline2026.0652.9o1-miniModel Setup=baselineModel Setup=baseline2026.0647.1