Web Research on Xbench DeepSearch
64.6Pass@1Claude-4-Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude-4-Sonnet2025.12 | 64.6 | |
| THINKMERGEmodel=WebSailor-32B, N=82025.12 | 57.6 | |
| THINKMERGEmodel=WebSailor-32B, N=42025.12 | 55.2 | |
| WebSailor-32Bvariant=Baseline2025.12 | 50.4 | |
| THINKMERGEmodel=WebSailor-32B, N=22025.12 | 50.2 | |
| THINKMERGEmodel=WebSailor-7B, N=42025.12 | 48 | |
| THINKMERGEmodel=WebSailor-7B, N=82025.12 | 47.2 | |
| THINKMERGEmodel=WebSailor-7B, N=22025.12 | 43.2 | |
| ASearcher-Web-QwQparameters=32B2025.12 | 42.1 | |
| WebSailor-7Bvariant=Baseline2025.12 | 37.8 | |
| FoldAct-7Bp_drop=0, parameters=7B2025.12 | 35.4 | |
| WebSailor-7Bparameters=7B2025.12 | 34.3 | |
| FoldAct-7Bp_drop=0.5, parameters=7B2025.12 | 32.9 | |
| THINKMERGEmodel=WebSailor-3B, N=22025.12 | 26.8 | |
| WebSailor-3Bvariant=Baseline2025.12 | 26.4 | |
| GPT-4.1-mini2025.12 | 25.4 | |
| THINKMERGEmodel=WebSailor-3B, N=42025.12 | 12.2 | |
| THINKMERGEmodel=WebSailor-3B, N=82025.12 | 5.4 |