Deep Research on GAIA (Pass@1, Pass@3)
70.5Pass@1OpenAI-o3
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenAI-o3Model Category=Advanced Models2025.10 | 70.5 | — | |
| Claude-4-SonnetModel Category=Advanced Models2025.10 | 68.3 | — | |
| DeepSeek-V3.1Model Category=Advanced Models2025.10 | 63.1 | — | |
| Kimi-K2Model Category=Advanced Models2025.10 | 57.7 | — | |
| WebSailorModel Size=32B2025.10 | 53.2 | — | |
| WebDancer-QwQ2025.10 | 51.5 | — | |
| WebThinker-RL2025.10 | 48.5 | — | |
| Web-30B-E-GRPOModel Size=30B, Training Strategy=E-GRPO2025.10 | 48.5 | 65.1 | |
| Web-30B-GRPOModel Size=30B, Training Strategy=GRPO2025.10 | 47.6 | 62.1 | |
| Web-30B-SFTModel Size=30B, Training Strategy=SFT2025.10 | 45 | 60.2 | |
| WebSailorModel Size=7B2025.10 | 37.9 | — | |
| Web-7B-E-GRPOModel Size=7B, Training Strategy=E-GRPO2025.10 | 36.9 | 51.5 | |
| Web-7B-GRPOModel Size=7B, Training Strategy=GRPO2025.10 | 33 | 44.7 | |
| Web-7B-SFTModel Size=7B, Training Strategy=SFT2025.10 | 31.7 | 44.7 | |
| R1-SearcherModel Size=7B2025.10 | 20.4 | — |