Google DeepMind launched Gemini Robotics 2 to enable integrated control over a robot's entire body. The new family of AI models advances beyond previous versions that focused primarily on upper-body movements.
Humanoid robots can now execute complex tasks by combining walking, crouching, and object manipulation. A single AI model manages these long-duration actions simultaneously. The update also introduces multi-robot collaboration for coordinating tasks across different machine types.
Gemini Robotics 2 includes specialized models for action, embodied reasoning, and on-device operations. These components enhance a robot's ability to plan and adapt to its environment. Google positions this release as a major step toward achieving general-purpose physical Artificial General Intelligence (AGI).