According to CoinMeta, Google has released a new generation of Gemini Robotics model family, which includes Gemini Robotics 2, ER 2, and On-Device 2. Gemini Robotics 2 is responsible for controlling actions, ER 2 is in charge of understanding the environment and planning tasks, while On-Device 2 can run directly on the robot's local system. This model is capable of converting visual images and text commands into actions, and a single model checkpoint can control humanoid robots, dual-arm robots, and different types of manipulators, eliminating the need to develop separate models for each type of robot. In addition, Gemini Robotics 2 has been enhanced with full-body control capabilities, enabling it to walk, squat, maintain balance, and perform precise movements. ER 2 is responsible for breaking down complex tasks, tracking progress, and coordinating the cooperation of multiple robots. Currently, ER 2 is available on Google AI Studio and has entered the private preview phase on Gemini Enterprise Agent Platform; Gemini Robotics 2 and On-Device 2 are still only available to early partners.