Google DeepMind’s Gemini Robotics 2 now enables robots to coordinate their full bodies, from torso to legs, allowing tasks like unscrewing a light bulb that require balance and precise force. The 2025 model previously controlled only a robot’s upper body. Carolina Parada, Google DeepMind’s vice president of robotics, said the goal is to bring AI into the physical world and build an intelligence layer usable by every robot. The system integrates vision, planning, and motor control into a single model. The demonstration used Apollo, a humanoid robot made by Apptronik Inc., with Agile Robots SE and Boston Dynamics as early-access partners. Alphabet is positioning itself as an operating system for physical AI, similar to Android. Google is opening a waitlist for developers and making three models available: Gemini Robotics ER 2 via Google’s AI Studio and a private preview on its enterprise platform, while Gemini Robotics 2 and On-Device 2 are shared with more than 100 trusted testers.