Agentic Reasoning on GAIA (val)
86.06Average ScoreInternAgent-1.5
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| InternAgent-1.5Base Model=Gemini-3-pro+o4-mini2026.02 | 86.06 | 92.45 | 89.53 | 61.54 | — | |
| MiroThinkerBase Model=MiroThinker-v1.5-235B2026.02 | 80.8 | — | — | — | — | |
| InternAgent-1.5Base Model=o4-mini2026.02 | 80.61 | 88.68 | 81.39 | 61.54 | — | |
| Manus2026.02 | 73.3 | 86.5 | 70.1 | 57.7 | — | |
| MiroThinkerBase Model=MiroThinker-v1.5-30B2026.02 | 72 | — | — | — | — | |
| Tongyi-DRBase Model=Tongyi-DR-30B2026.02 | 70.9 | — | — | — | — | |
| OWLBase Model=Gemini 2.5 Pro2026.02 | 69.7 | 84.9 | 68.6 | 42.3 | — | |
| OpenAI DR2026.02 | 67.36 | 74.29 | 69.06 | 47.6 | — | |
| WebShaperBase Model=Qwen2.5-72B2026.02 | 60.1 | 69.2 | 63.4 | 16.6 | — | |
| InternAgent-1.5Base Model=Qwen3-235B2026.02 | 58.79 | 69.81 | 60.47 | 30.77 | — | |
| WebDancerBase Model=QwQ-32B2026.02 | 51.5 | 61.5 | 50 | 25 | — | |
| Learning to ShareShared Memory=true, Model=GPT-5.12026.02 | 49.1 | 62.3 | 47.7 | 26.9 | 781 | |
| M1-ParallelShared Memory=false, Model=GPT-5.12026.02 | 47.9 | 60.4 | 46.5 | 26.9 | 1,005 | |
| MagenticOneShared Memory=false, Model=GPT-5.12026.02 | 37.5 | 48.1 | 37.3 | 16.7 | 815 | |
| Learning to ShareShared Memory=true, Model=Qwen3 (32B)2026.02 | 30.4 | 47.2 | 25.6 | 11.8 | 892 | |
| M1-ParallelShared Memory=false, Model=Qwen3 (32B)2026.02 | 24.8 | 43.4 | 18.6 | 7.7 | 1,569 | |
| MagenticOneShared Memory=false, Model=Qwen3 (32B)2026.02 | 22.5 | 41.5 | 16.3 | 3.8 | 758 |