Google reveals Gemini Robotics 2.0, promising improved dexterity and safety
Ars Technica AI · Ryan Whitwam · 2026-07-30
Google DeepMind releases Gemini Robotics 2.0, a trio of AI sub-models enabling humanoid robots to perform more complex tasks with improved dexterity, multi-robot collaboration, and continuous environmental analysis, with one sub-model immediately available to developers.
Appears in
Extraction
Topics: roboticsembodied-aigoogle-deepmindgeneralist-robots
Claims
- Gemini Robotics 2.0 comprises three new sub-models, one of which—Gemini Robotics ER 2—is available to developers via the Gemini Live API starting today.
- Gemini Robotics ER 2 is an upgraded embodied reasoning model that DeepMind claims represents a significant leap over the previous 1.6 release.
- The 2.0 release brings 'whole-body intelligence,' enabling AI control of complete humanoid robots including complex hands with improved dexterity.
- Google DeepMind's stated goal is a generalist robot capable of executing arbitrary human-directed tasks, which its scientists describe as 'physical AGI.'
Key quotes
The goal of Gemini robotics is to create a generalist robot, one that can do anything a human could do. Google DeepMind scientists sometimes call this 'physical AGI.'
With the 2.0 release, Google says its robotics AI can control an entire humanoid robot with improved dexterity, even for machines with complex humanoid hands.