Google Deepmind has released Gemini Robotics 2, an updated vision-language-action model designed to control physical robots ranging from small tabletop arms to full-body humanoids. This system acts as an intelligence layer that manages full-body movement, executes fine motor tasks, and coordinates multiple machines simultaneously. Developers can currently apply for early access through a waitlist. The company also launched Gemini Robotics ER 2, a newer embodied reasoning model that replaces the previous ER 1.6 version available in Google AI Studio. ER 2 focuses on understanding the physical environment to decide which actions to take next.
The significance lies in the shift toward adaptive hardware rather than fixed automation. By combining image recognition, language processing, and direct action control, these models allow software to guide diverse mechanical forms without extensive reprogramming for each specific task. This approach lowers the barrier for integrating AI into varied industrial and domestic settings where robots must operate in unstructured spaces.
