Egocentric daily-task planning on EgoPlanBench2
56.7Overall Success RateLoHo-Manip-4B
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LoHo-Manip-4BModel Category=Embodied Foundation Models, Parameters=4B2026.04 | 56.7 | 60.4 | 54.8 | 51.9 | 56.4 | |
| Gemini-3.0-FlashModel Category=Proprietary Models (API)2026.04 | 48.8 | 55.8 | 46.3 | 45.9 | 45.7 | |
| ThinkAct-7BModel Size=7B2026.01 | 48.2 | 50.1 | 45.2 | 44.8 | 49.8 | |
| ThinkAct-7BModel Category=Embodied Foundation Models, Parameters=7B2026.04 | 48.2 | 50.1 | 45.2 | 44.8 | 49.8 | |
| Fast-ThinkAct-7BModel Size=7B2026.01 | 47.5 | 51.3 | 45.9 | 41.5 | 47.3 | |
| Fast-ThinkAct-3BZero-shot=true2026.01 | 46.4 | 50.3 | 43.2 | 46.4 | 44.3 | |
| Fast-ThinkAct-3BModel Category=Embodied Foundation Models, Parameters=3B2026.04 | 46.4 | 50.3 | 43.2 | 46.4 | 44.3 | |
| ThinkAct-3BZero-shot=true2026.01 | 44 | 46.6 | 42.5 | 45.9 | 41.4 | |
| ThinkAct-3BModel Category=Embodied Foundation Models, Parameters=3B2026.04 | 44 | 46.6 | 42.5 | 45.9 | 41.4 | |
| Eagle-2.5-8BAccess=Open-Source, Parameters=8B2026.02 | 43 | 49.4 | 41.7 | 47 | 35.1 | |
| Gemini-2.5-ProAccess=Closed-Source2026.02 | 42.8 | 44.2 | 43.1 | 46.4 | 39.6 | |
| Gemini-2.5-FlashZero-shot=true2026.01 | 42.4 | 44.2 | 39.1 | 43.2 | 42.3 | |
| Gemini-2.5-FlashModel Category=Proprietary Models (API)2026.04 | 42.4 | 44.2 | 39.1 | 43.2 | 42.3 | |
| PhyCritic-7BAccess=Open-Source, Parameters=7B2026.02 | 42.3 | 45.6 | 38.3 | 47.5 | 39.4 | |
| GPT-4oAccess=Closed-Source2026.02 | 41.8 | 47.4 | 40 | 44.8 | 35.6 | |
| RoboBrain2.0-3BZero-shot=true2026.01 | 41.8 | 45.3 | 39.7 | 45.9 | 37.6 | |
| RoboBrain2.0-3BModel Category=Embodied Foundation Models, Parameters=3B2026.04 | 41.8 | 45.3 | 39.7 | 45.9 | 37.6 | |
| RoboBrain2.5-8BModel Category=Embodied Foundation Models, Parameters=8B2026.04 | 41.2 | 43.5 | 40.1 | 38.8 | 40.8 | |
| Gemini-2.5-FlashAccess=Closed-Source2026.02 | 37.1 | 38.7 | 35.6 | 43.7 | 33.4 | |
| Qwen3-VL-8BModel Category=Open-source Models, Parameters=8B2026.04 | 36.6 | 39.6 | 31.6 | 38.8 | 37.7 | |
| InternVL3-8BModel Size=8B2026.01 | 36.2 | 38.5 | 37.2 | 36.1 | 32.9 | |
| RynnBrain-8BModel Category=Embodied Foundation Models, Parameters=8B2026.04 | 34.8 | 38.9 | 31.4 | 35.5 | 32.6 | |
| NVILA-8BModel Size=8B2026.01 | 33.7 | 35.8 | 35.4 | 37.2 | 28.7 | |
| InternVL2.5-8BModel Size=8B2026.01 | 33.5 | 36.2 | 35.4 | 34.4 | 28.7 | |
| InternVL3-2BZero-shot=true2026.01 | 33.4 | 36.9 | 31.5 | 35.6 | 29.9 | |
| InternVL3-2BModel Category=Open-source Models, Parameters=2B2026.04 | 33.4 | 36.9 | 31.5 | 35.6 | 29.9 | |
| Robobrain-2.0-7BAccess=Open-Source, Parameters=7B2026.02 | 33.2 | 39.4 | 32.2 | 33.9 | 29.7 | |
| RoboBrain2.0-7BModel Size=7B2026.01 | 33.2 | 39.4 | 32.2 | 33.9 | 27 | |
| RoboBrain2.0-7BModel Category=Embodied Foundation Models, Parameters=7B2026.04 | 33.2 | 39.4 | 32.2 | 33.9 | 29.7 | |
| Qwen2.5-VL-7BAccess=Open-Source, Parameters=7B2026.02 | 32.6 | 36.7 | 27.7 | 33.9 | 32.5 | |
| GPT-4VZero-shot=true2026.01 | 32.6 | 36.7 | 32.5 | 33.9 | 27.7 | |
| GPT-4VModel Category=Proprietary Models (API)2026.04 | 32.6 | 36.7 | 32.5 | 33.9 | 27.7 | |
| NVILA-2BZero-shot=true2026.01 | 31.4 | 34.6 | 31.6 | 33.3 | 26.7 | |
| NVILA-2BModel Category=Open-source Models, Parameters=2B2026.04 | 31.4 | 34.6 | 31.6 | 33.3 | 26.7 | |
| Qwen3-VL-4BModel Category=Open-source Models, Parameters=4B2026.04 | 31.3 | 32.3 | 28.9 | 33.8 | 31.2 | |
| InternVL2.5-2BZero-shot=true2026.01 | 30.1 | 30.9 | 33.1 | 28.6 | 27.8 | |
| InternVL2.5-2BModel Category=Open-source Models, Parameters=2B2026.04 | 30.1 | 30.9 | 33.1 | 28.6 | 27.8 | |
| Cosmos-R1-7BAccess=Open-Source, Parameters=7B2026.02 | 29.8 | 33.5 | 31.2 | 30.6 | 24.5 | |
| Magma-8BZero-shot=true2026.01 | 29.8 | 32.1 | 29.3 | 34.4 | 25.7 | |
| Magma-8BModel Size=8B2026.01 | 29.8 | 32.1 | 29.3 | 34.4 | 25.7 | |
| Magma-8BModel Category=Embodied Foundation Models, Parameters=8B2026.04 | 29.8 | 32.1 | 29.3 | 34.4 | 25.7 | |
| Qwen2.5-VL-7BModel Size=7B2026.01 | 29.1 | 31.4 | 28.6 | 29.5 | 26.7 | |
| Qwen2.5-VL-3BZero-shot=true2026.01 | 28.5 | 29 | 28.9 | 30.2 | 27 | |
| Qwen2.5-VL-3BModel Category=Open-source Models, Parameters=3B2026.04 | 28.5 | 29 | 28.9 | 30.2 | 27 |