Gemini Robotics: DeepMind brings foundation models into the physical world
In one sentence Google DeepMind updates Gemini Robotics and Gemini Robotics-ER: generalist VLAs on Gemini 2 base that drive industrial arms and humanoids (Apptronik Apollo) zero-shot on never-seen tasks.
DeepMind updates its two robotics model families: Gemini Robotics (end-to-end action, the output is the movement) and Gemini Robotics-ER (an "Embodied Reasoning" version that thinks spatially before delegating execution to a controller).
The 2025 key novelty: these models are no longer tied to a single robot. The same model drives a lab industrial arm, an Apptronik Apollo humanoid, even ALOHA robots with different designs. The phrase is "cross-embodiment generalist VLA".
Strategic message: DeepMind, which historically doesn't build robot hardware, bets on a universal model that becomes the "base brain" for all robots. Apptronik becomes the privileged partner for industrial validation (warehouse, manufacturing).
Companies
Google DeepMind, Apptronik
Tools
Gemini Robotics, Gemini Robotics-ER
Tags
Sources