Multi-turn tool use on BrowseComp-ZH
58.1Pass@1OpenAI-o3†
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenAI-o3†RL Algo=-2026.01 | 58.1 | — | |
| DeepSeek-V3.1†RL Algo=-2026.01 | 49.2 | — | |
| WebAnchor-30BRL Algo=Anchor-GRPO2026.01 | 48.8 | 65.1 | |
| WebAnchor-30BRL Algo=First-step GRPO2026.01 | 47.9 | 63.2 | |
| WebAnchor-30BRL Algo=GRPO2026.01 | 46.2 | 61.4 | |
| Claude-4-Sonnet†RL Algo=-2026.01 | 29.1 | — | |
| Kimi-K2†RL Algo=-2026.01 | 28.8 | — | |
| WebSailor-32BRL Algo=-2026.01 | 25.5 | — | |
| WebAnchor-7BRL Algo=Anchor-GRPO2026.01 | 23.8 | 36.7 | |
| WebDancer-QwQRL Algo=-2026.01 | 18 | — | |
| WebAnchor-7BRL Algo=GRPO2026.01 | 17.5 | 31.5 | |
| WebAnchor-7BRL Algo=First-step GRPO2026.01 | 16.4 | 31.2 | |
| WebSailor-7BRL Algo=-2026.01 | 14.2 | — | |
| WebAnchor-3BRL Algo=Anchor-GRPO2026.01 | 13.1 | 25 | |
| WebAnchor-3BRL Algo=First-step GRPO2026.01 | 12.9 | 21.5 | |
| WebAnchor-3BRL Algo=GRPO2026.01 | 12.3 | 19.2 | |
| WebThinker-RLRL Algo=-2026.01 | 7.3 | — | |
| R1-Searcher-7BRL Algo=-2026.01 | 0.6 | — |