Deep Search and Information Retrieval on xbench DeepSearch 2510
75Avg@8OpenAI-GPT-5
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OpenAI-GPT-5Agent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 75 | — | — | |
| MiroThinker-H1Agent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 72 | — | — | |
| MiroThinker-1.7Agent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 62 | — | — | |
| MiroThinker-1.7-miniAgent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 57.2 | — | — | |
| Tongyi-DeepResearch-30BAgent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 55 | — | — | |
| Gemini-3.0-ProAgent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 53 | — | — | |
| Kimi-K2.5Agent Style=ReAct, Max interaction turns=200, Max episode retries=5, Context management retention=52026.03 | 46 | — | — | |
| FORT-SearcherAgent Category=Comparable-size Open-source Agents2026.06 | — | 57.2 | 66.2 | |
| Kimi-K2.5-ThinkingAgent Category=Large-scale Open-source Agents2026.06 | — | 57.4 | — | |
| MiroThinker-1.7-miniAgent Category=Comparable-size Open-source Agents2026.06 | — | 57.2 | 64.6 | |
| OpenSeekerV2Agent Category=Comparable-size Open-source Agents2026.06 | — | 43.4 | 53.4 | |
| Qwen3.5-35B-A3BAgent Category=Comparable-size Open-source Agents2026.06 | — | 50.3 | 59.9 | |
| Step 3.5 FlashAgent Category=Large-scale Open-source Agents2026.06 | — | 56.3 | — | |
| Tongyi DeepResearchAgent Category=Comparable-size Open-source Agents2026.06 | — | 47.5 | 51.7 |