Multi-turn tool use on Xbench
75.1Pass@1WebAnchor-30B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| WebAnchor-30BRL Algo=Anchor-GRPO2026.01 | 75.1 | 85 | |
| WebAnchor-30BRL Algo=First-step GRPO2026.01 | 72 | 88 | |
| DeepSeek-V3.1†RL Algo=-2026.01 | 71 | — | |
| WebAnchor-30BRL Algo=GRPO2026.01 | 71 | 84 | |
| OpenAI-o3†RL Algo=-2026.01 | 66.7 | — | |
| Claude-4-Sonnet†RL Algo=-2026.01 | 64.6 | — | |
| WebSailor-32BRL Algo=-2026.01 | 53.3 | — | |
| Kimi-K2†RL Algo=-2026.01 | 50 | — | |
| WebAnchor-7BRL Algo=Anchor-GRPO2026.01 | 45 | 66 | |
| WebAnchor-7BRL Algo=GRPO2026.01 | 40.7 | 56 | |
| WebDancer-QwQRL Algo=-2026.01 | 39 | — | |
| WebAnchor-3BRL Algo=Anchor-GRPO2026.01 | 37.7 | 56 | |
| WebAnchor-7BRL Algo=First-step GRPO2026.01 | 37.7 | 52 | |
| WebAnchor-3BRL Algo=First-step GRPO2026.01 | 34.7 | 55 | |
| WebSailor-7BRL Algo=-2026.01 | 34.3 | — | |
| WebAnchor-3BRL Algo=GRPO2026.01 | 31.6 | 55 | |
| WebThinker-RLRL Algo=-2026.01 | 24 | — | |
| R1-Searcher-7BRL Algo=-2026.01 | 4 | — |