Multi-turn tool use on BrowseComp
50.9Pass@1OpenAI-o3†
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenAI-o3†RL Algo=-2026.01 | 50.9 | — | |
| WebAnchor-30BRL Algo=Anchor-GRPO2026.01 | 46 | 67.5 | |
| WebAnchor-30BRL Algo=GRPO2026.01 | 44 | 65 | |
| WebAnchor-30BRL Algo=First-step GRPO2026.01 | 43.3 | 66 | |
| DeepSeek-V3.1†RL Algo=-2026.01 | 30 | — | |
| Kimi-K2†RL Algo=-2026.01 | 14.1 | — | |
| Claude-4-Sonnet†RL Algo=-2026.01 | 12.2 | — | |
| WebAnchor-7BRL Algo=Anchor-GRPO2026.01 | 11.6 | 20.7 | |
| WebSailor-32BRL Algo=-2026.01 | 10.5 | — | |
| WebAnchor-7BRL Algo=First-step GRPO2026.01 | 9.3 | 16.3 | |
| WebAnchor-3BRL Algo=Anchor-GRPO2026.01 | 7.1 | 11.5 | |
| WebAnchor-3BRL Algo=First-step GRPO2026.01 | 6.9 | 10.2 | |
| WebSailor-7BRL Algo=-2026.01 | 6.7 | — | |
| WebAnchor-7BRL Algo=GRPO2026.01 | 6.3 | 11.7 | |
| WebAnchor-3BRL Algo=GRPO2026.01 | 5.1 | 8.7 | |
| WebDancer-QwQRL Algo=-2026.01 | 3.8 | — | |
| WebThinker-RLRL Algo=-2026.01 | 2.8 | — | |
| R1-Searcher-7BRL Algo=-2026.01 | 0.4 | — |