Gemini Robotics On-Device 2
READER SUMMARYAn efficient local vision-language-action model that adapts to new robot bodies from fewer than 200 examples and operates without continuous network access.
Why this model scores 60.2
Local operation and rapid embodiment adaptation increase autonomy and make centralized monitoring harder, but the model is less capable than the larger Robotics 2 tiers and remains limited to early-access partners. Physical safety frameworks and constrained access reduce deployment and misuse, while on-device execution keeps residual control difficulty elevated.
News tied to Gemini Robotics On-Device 2
The model score of 60.2 rates this model's risk profile. The overall Doom Index of 67.7 measures the accumulated evidence trajectory. These values answer different questions.
Each article's signed Doom Index movement is divided equally among the exact models named on that article. This prevents multi-model evidence from being counted in full on several model pages. The model's risk score itself never changes the overall index.
Google DeepMind releases Gemini Robotics 2 for whole-body and multi-robot tasks
Three new robotics models demonstrated whole-body humanoid control, dexterous manipulation, several-minute planning, self-correction, multi-robot collaboration, and rapid on-device adaptation, with reasoning access available in AI Studio.
- Full item movement
- +3.27
- Gemini Robotics On-Device 2 equal share
- +1.09
Model score history
- R1Doom Score 60.2
Initial source-backed profile for the exact on-device release.
11 Aug 2026