Deep Search on xbench DeepSearch (test)
75AccuracyTongyi-DeepResearch
Evaluation results
| Method | Links | |
|---|---|---|
| Tongyi-DeepResearchContext Budget=128k2026.01 | 75 | |
| DeepSeek-v3.1Context Budget=128k2026.01 | 71.2 | |
| GLM-4.5Context Budget=128k2026.01 | 70 | |
| OpenAI-o3Context Budget=128k2026.01 | 66.7 | |
| DeepDive-30B-SFT + C-GRPOContext Budget=128k2026.01 | 57.7 | |
| DeepDive-30B-SFT + E-GRPOContext Budget=128k2026.01 | 55.7 | |
| DeepDive-30B-SFT + C-GRPOContext Budget=64k2026.01 | 55.3 | |
| DeepDive-30B-SFTContext Budget=128k2026.01 | 54.3 | |
| DeepDive-4B-SFT + C-GRPOContext Budget=128k2026.01 | 54 | |
| WebExplorer-8BContext Budget=128k2026.01 | 53.7 | |
| WebSailor-32BContext Budget=32k2026.01 | 53.3 | |
| DeepDive-30B-SFT + GRPOContext Budget=128k2026.01 | 52 | |
| DeepDive-9BContext Budget=128k2026.01 | 51.8 | |
| DeepDive-32BContext Budget=128k2026.01 | 51.8 | |
| DeepDive-30B-SFT + E-GRPOContext Budget=64k2026.01 | 51.7 | |
| DeepDive-30B-SFT + GRPOContext Budget=64k2026.01 | 51.3 | |
| DeepDive-4B-SFT + C-GRPOContext Budget=64k2026.01 | 50.3 | |
| DeepDive-4B-SFT + E-GRPOContext Budget=128k2026.01 | 45 | |
| DeepDive-4B-SFTContext Budget=128k2026.01 | 44.3 | |
| DeepDive-4B-SFT + E-GRPOContext Budget=64k2026.01 | 43.7 | |
| DeepDive-30B-SFTContext Budget=64k2026.01 | 43 | |
| Asearcher-Web-32BContext Budget=128k2026.01 | 42.1 | |
| DeepDive-4B-SFT + GRPOContext Budget=128k2026.01 | 41.3 | |
| DeepDive-4B-SFT + GRPOContext Budget=64k2026.01 | 41 | |
| WebSailor-7BContext Budget=32k2026.01 | 34.3 | |
| DeepDive-4B-SFTContext Budget=64k2026.01 | 34 |