A major step for humanoid robot intelligence
Google DeepMind has introduced Gemini Robotics 2, a new AI model designed to control an entire humanoid robot body. Unlike the previous version, which focused on upper-body movement, the new model coordinates motion from a robot’s feet all the way to its fingertips.
That broader control enables a much wider range of useful actions. Google says robots powered by Gemini Robotics 2 can walk, crouch, stretch, reach, and manipulate objects, helping them navigate environments built for people.
In demonstrations shared by Google, Apptronik’s Apollo 2 robot bends down to pick up a watering can and identifies specific items on a shelf before retrieving them. These examples show how whole-body AI control can make humanoid robots more adaptable and practical in everyday settings.
- More mobility: Robots can coordinate legs, torso, arms, and hands together.
- More useful tasks: Whole-body movement supports actions like reaching low shelves or handling objects while balancing.
- Path to real-world helpers: The technology could accelerate progress toward robots that assist in homes, workplaces, warehouses, and care settings.
While this is still an early-stage robotics milestone, it is a promising sign of how AI models are beginning to bridge perception, reasoning, and physical action. Gemini Robotics 2 brings humanoid robots closer to operating smoothly in the real world.