Autonomous Web Navigation on WebArena 812 tasks
71.2Success RateColorBrowserAgent
Evaluation Results
| Method | Links | |
|---|---|---|
| ColorBrowserAgentOpen-Source=Yes, Backbone=GPT-5, Temperature=0.02026.01 | 71.2 | |
| Claude Code + GBOX MCPOpen-Source=Yes2026.01 | 68 | |
| DeepSky AgentOpen-Source=No2026.01 | 66.9 | |
| IBM CUGAOpen-Source=Yes2026.01 | 61.7 | |
| OpenAI OperatorOpen-Source=No2026.01 | 58.1 | |
| WebOperatorOpen-Source=Yes2026.01 | 54.6 | |
| AgentSymbioticOpen-Source=Yes2026.01 | 52.1 |