Gemini Robotics and Gemini Robotics-ER
Gemini Robotics, a vision-language-action model built on Gemini 2.0, plus Gemini Robotics-ER for spatial reasoning; reported more than double prior VLAs on generalization.
Adds physical actions as an output modality to Gemini 2.0 so one model can directly control varied robots. The ER variant exposes spatial and embodied reasoning for developers to drive their own robot controllers. Apptronik is the humanoid partner.
- Date
- Wednesday, 12 March 2025
- Lab
- Google DeepMind
- Kind
- model
- Access
- research preview
Figures
| Measure | Value | Measured by |
|---|---|---|
| Generalization benchmark vs other SOTA VLAs | >2x on average per Google tech report (company-measured) | company |
ER was available to trusted testers (Agile Robots, Agility Robots, Boston Dynamics, Enchanted Tools). Successors were On-Device (2025-06-24), 1.5 (2025-09-25) and Robotics 2 (2026-07-30).
Sources
- deepmind.google/blog/gemini-robotics-brings-ai-into-the-physical-world/
- blog.google/technology/ai/google-ai-updates-march-2025/
This record was checked against its sources on 6 October 2026. How we check