General Embodied AI Performance on Aggregate Sokoban, FrozenLake, Navigation, PrimitiveSkill
93Overall Success RateAtlasVA
Evaluation Results
| Method | Links | |
|---|---|---|
| AtlasVAModel Category=Ours, Parameter Count=3B2026.05 | 93 | |
| VAGENModel Category=Open-Source2026.05 | 78 | |
| o3Model Category=Proprietary2026.05 | 71 | |
| GPT-5Model Category=Proprietary2026.05 | 69 | |
| Claude Sonnet 4.5Model Category=Proprietary2026.05 | 62 | |
| o4-miniModel Category=Proprietary2026.05 | 60 | |
| GPT-4oModel Category=Proprietary2026.05 | 60 | |
| Gemini 2.5 flashModel Category=Proprietary2026.05 | 58 | |
| Qwen2.5-VL-72BModel Category=Open-Source2026.05 | 55 | |
| Gemini 2.5 ProModel Category=Proprietary2026.05 | 51 | |
| Claude Sonnet 3.7Model Category=Proprietary2026.05 | 51 | |
| Gemini 2.0Model Category=Proprietary2026.05 | 39 | |
| Qwen2.5-VL-7BModel Category=Open-Source2026.05 | 19 | |
| VLM-R1-3BModel Category=Open-Source2026.05 | 10 | |
| Qwen2.5-VL-3BModel Category=Open-Source2026.05 | 9 |