Long-horizon agentic tasks on BrowseComp-ZH Our Settings
71.3Pass@1AgentSwing
Evaluation Results
| Method | Links | |
|---|---|---|
| AgentSwingBase Model=DeepSeek-v3.22026.03 | 71.3 | |
| DeepSeek-v3.2Context Management=Discard-All2026.03 | 70.2 | |
| DeepSeek-v3.2Context Management=Keep-Last-N2026.03 | 69.9 | |
| DeepSeek-v3.2Context Management=Summary2026.03 | 69.2 | |
| Gemini-3.0-ProContext Management=w/o CM2026.03 | 66.8 | |
| MiroThinker-v1.5-30B-A3BContext Management=-2026.03 | 66.8 | |
| Claude-4.5-OpusContext Management=w/o CM2026.03 | 62.4 | |
| DeepSeek-v3.2Context Management=Baseline (w/o CM)2026.03 | 61.6 | |
| OpenAI-o3Context Management=w/o CM2026.03 | 58.1 | |
| AgentSwingBase Model=Tongyi-DR-30B-A3B2026.03 | 56.7 | |
| Tongyi-DR-30B-A3BContext Management=Discard-All2026.03 | 53.9 | |
| Tongyi-DR-30B-A3BContext Management=Keep-Last-N2026.03 | 50.1 | |
| Tongyi-DR-30B-A3BContext Management=Summary2026.03 | 49.1 | |
| AgentFold-30B-A3BContext Management=-2026.03 | 47.3 | |
| Tongyi-DR-30B-A3BContext Management=Baseline (w/o CM)2026.03 | 47.1 | |
| IterResearch-30B-A3BContext Management=-2026.03 | 45.2 | |
| AgentFounder-30B-A3BContext Management=-2026.03 | 43.3 | |
| OpenAI DeepResearchContext Management=-2026.03 | 42.9 | |
| DeepMiner-32B-RLContext Management=-2026.03 | 40.1 | |
| AgentSwingBase Model=GPT-OSS-120B2026.03 | 38 | |
| GPT-OSS-120BContext Management=Keep-Last-N2026.03 | 33.6 | |
| GPT-OSS-120BContext Management=Discard-All2026.03 | 31.5 | |
| GPT-OSS-120BContext Management=Summary2026.03 | 30.8 | |
| GPT-OSS-120BContext Management=Baseline (w/o CM)2026.03 | 28.4 | |
| ASearcher-Web-32BContext Management=-2026.03 | 15.6 |