Google DeepMind has released a new generation of robotic AI models, Gemini Robotics 2, which enables full-body control of humanoid robots from head to toe. It also simultaneously launched two additional robotic AI models—Gemini Robotics ER 2 and On-Device 2—which can operate collaboratively or independently, supporting multi-step task planning and coordination among multiple robots. In a pre-recorded demonstration for the media on Tuesday, the model directed Apptronik’s Apollo humanoid robot through a complete sequence: navigating across a room, picking up a watering can, and placing it on a lower shelf—all while autonomously avoiding obstacles. As of this report, Alphabet’s stock was down 1.4% during Thursday’s trading session. Carolina Parada, Vice President of Robotics at Google DeepMind, stated at a press briefing: "Our goal is to bring AI into the physical world and build an intelligent layer that every robot can use." However, Kanishka Rao, Director of Robotics at Google DeepMind, acknowledged that genuine dexterity remains a distant objective. Robotic movements remain slow and deliberate, as machines must pause to deliberate over decisions that humans make intuitively. This launch continues Google's robotics strategy for the year. In 2025, Google introduced the first-generation Gemini Robotics, which can translate language and visual information into robotic actions, reigniting the company’s robotics ambitions that have spanned over a decade. After acquiring several robotics startups in the early 201