Gemini Robotics 2 brings whole body intelligence to robots (deepmind.google)

Gemini Robotics has introduced Gemini Robotics 2, a groundbreaking intelligence layer that enhances robots' abilities in whole-body control, dexterity, and teamwork. Traditional robots are limited to narrow tasks and struggle with adaptability, but this new platform equips robots with the capability to learn autonomously and handle unpredictable environments. With Gemini Robotics 2, robots can execute complex movements—such as walking, crouching, and object manipulation—while collaborating with other robots to complete jobs efficiently. The significant technical advancements of Gemini Robotics 2 stem from three core models: the Gemini Robotics 2, which merges vision and language for intricate motor control; Gemini Robotics ER 2, an embodied reasoning model that enables multi-step planning and communication with humans; and Gemini Robotics On-Device 2, optimized for local execution that allows rapid adaptation to different robotic forms. These innovations not only enhance manipulation and reasoning but also set a new standard for the integration of AI in robotics, making the dream of versatile robots that can interact seamlessly in human environments a closer reality.