Google DeepMind has made available three AI models designed for robots. The models, part of the Gemini Robotics 2 intelligence layer, enable whole body control, dexterity, and multi-robot collaboration. A vision-language-action model allows for humanoid control, while an embodied reasoning model coordinates tasks. A third model adapts to new robot bodies quickly. Only one of the models is currently publicly accessible. The release marks an advancement in AI capabilities for robots. This development is significant as it brings robots closer to achieving human-like dexterity and coordination.