Web Browsing and Task Completion on BrowseComp
58.3Pass@1 RateTongyi DeepResearch
Evaluation Results
| Method | Links | |
|---|---|---|
| Tongyi DeepResearchAgent Type=DeepResearch Agent, Mode=Heavy Mode2025.10 | 58.3 | |
| OpenAI DeepResearchAgent Type=DeepResearch Agent2025.10 | 51.5 | |
| OpenAI o3Agent Type=LLM-based ReAct Agent2025.10 | 49.7 | |
| Tongyi DeepResearchAgent Type=DeepResearch Agent, Configuration=30B-A3B2025.10 | 43.4 | |
| DeepSeek-V3.1Agent Type=LLM-based ReAct Agent2025.10 | 30 | |
| OpenAI o4-miniAgent Type=LLM-based ReAct Agent2025.10 | 28.3 | |
| GLM 4.5Agent Type=LLM-based ReAct Agent2025.10 | 26.4 | |
| Kimi K2Agent Type=LLM-based ReAct Agent2025.10 | 14.1 | |
| Claude-4-SonnetAgent Type=LLM-based ReAct Agent2025.10 | 12.2 |