Model / Google DeepMind

Gemini Robotics 2

Vision-language-action model for robot control across embodiments.

Embodiment role

Google DeepMind describes Gemini Robotics 2 as a VLA model for controlling different robot embodiments, including humanoids, with capabilities spanning dexterity, whole-body control and multi-robot collaboration.

Related layers

Research · Robots · Companies · Glossary

Primary source

Google DeepMind — Official source →

Editorial model profile · Reviewed September 2026