Google wants AI to do more than answer questions on a screen. It wants AI to work in the real world. On March 12, 2025, Google DeepMind introduced Gemini Robotics and Gemini Robotics-ER, two models built for robots. Google said these models can use language, images, and an understanding of the physical world to help robots act. The company also said the system could do careful jobs such as folding paper, packing items, and working with different kinds of robots, including Apptronik’s humanoid robot Apollo. (deepmind.google)
Then Google made the idea even more practical. On June 24, 2025, it announced Gemini Robotics On-Device. This version can run directly on a robot, so it does not need a strong network connection all the time. That can help robots react faster and keep working in places with weak or no internet. Google also said developers can adapt the model to new tasks with only about 50 to 100 examples, which could make training easier. (deepmind.google)
Another big step came on September 25, 2025. Google introduced Gemini Robotics 1.5 and Gemini Robotics-ER 1.5. Google explained that one model turns what the robot sees and hears into movement, while the ER model plans the job, reasons about the space around it, and can even use tools such as Google Search. In simple words, the robot can start to “see,” “think,” and “move” as one system. Google said this could help robots do longer, harder tasks, like sorting trash by local rules or cleaning a kitchen step by step. (deepmind.google)
As of April 15, 2026, Google’s newest robotics update is Gemini Robotics-ER 1.6. Google says it improves understanding of space, task planning, and success checking. It can also read complex gauges and sight glasses, a skill found through work with Boston Dynamics. Google calls it its safest robotics model so far, and it is now available through the Gemini API and Google AI Studio. Home robots that do everything are not here yet. But the age of robots that can look, think, and act together is getting much closer. (blog.google)










