Google DeepMind’s new AI model can control a robot’s entire body

Google DeepMind launched Gemini Robotics 2, expanding control from upper-body motion to full-body actions.

The system can now handle more complex tasks such as walking, crouching, stretching, and object manipulation.

It also improves hand dexterity and upgrades embodied reasoning for longer, multi-robot tasks.

An updated on-device model helps robots adapt faster to new designs with more local execution.