A Google DeepMind AI model family for controlling robots.
It is like sending a robot arm to diner cook school. It sees the pancake, hears “add an egg,” and does not launch breakfast.
It helps robot arms, helper robots, and local devices. It makes physical AI move better and follow orders.
Gemini
Gemini Robotics 2 extends Gemini multimodal skills into robot control.
VLA
Its VLA models turn vision and language into actions.
Embodied AI
Gemini Robotics 2 helps models understand and act in the physical world.
Edge AI
The local version can run on robot devices with limited network access.