Gemini Robotics

Gemini Robotics is an AI-powered system from Google DeepMind designed to control and teach robots through multimodal understanding and natural language. It combines vision, language, and robotics to interpret camera input, reason about environments, and generate precise, executable robot actions. The model can follow high-level verbal instructions, translate them into low-level control commands, and adapt to new tasks with minimal additional programming.

Key capabilities include object recognition, spatial reasoning, and task planning in real-world, unstructured environments such as homes, labs, and warehouses. Gemini Robotics supports learning from demonstration, allowing robots to generalize from a small number of examples and apply learned skills to related tasks. It can also explain its plans and actions in natural language, improving transparency and human-robot collaboration.