Web Browsing/Site Navigation on AgentBench WS
52Task Completion Score (TCS)Gemini-3-flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-3-flashAverage Steps (Stp)=12.6, Input Tokens (Tok In)=77.0, Output Tokens (Tok Out)=5.7, Execution Time=79.92026.05 | 52 | |
| Qwen3-235B-A22BAverage Steps (Stp)=4.8, Input Tokens (Tok In)=13.9, Output Tokens (Tok Out)=0.2, Execution Time=8.92026.05 | 40.6 | |
| GPT-5-miniAverage Steps (Stp)=16.6, Input Tokens (Tok In)=97.7, Output Tokens (Tok Out)=7.9, Execution Time=126.22026.05 | 40.4 | |
| DeepSeek-V3.2Average Steps (Stp)=16.4, Input Tokens (Tok In)=5.5, Output Tokens (Tok Out)=0.9, Execution Time=35.72026.05 | 32 |