Agentic Web Interaction on BrowseComp-EN (test)
61.5Pass@1GPT-5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-5Agent Category=Proprietary Agents2026.04 | 61.5 | — | |
| DeepSeek-V3.2Agent Category=Proprietary Agents2026.04 | 35 | — | |
| GLM-4.6Agent Category=Proprietary Agents2026.04 | 34.9 | — | |
| DeepSeek-V3.1Agent Category=Proprietary Agents2026.04 | 23.6 | — | |
| LThinker++Agent Category=Our Agents, Protocol=Behavioral Pruning constraints2026.04 | 18.1 | 31.5 | |
| Vanilla-AgentAgent Category=Our Agents, Protocol=+ SFT (vanilla)2026.04 | 16 | 27.3 | |
| Kimi-K2-InstructAgent Category=Proprietary Agents2026.04 | 14.1 | — | |
| Claude-4-SonnetAgent Category=Proprietary Agents2026.04 | 12.2 | — | |
| Qwen3-30B-A3B-ThinkingAgent Category=Our Agents, Protocol=Base2026.04 | 2.1 | 4 |