Google DeepMind Launches Gemini Robotics 2 for Whole-Body Control

Google DeepMind has unveiled Gemini Robotics 2, an advanced vision-language-action (VLA) model designed to provide whole-body intelligence to robots. This model enables sophisticated control of humanoids and other bi-arm robots, including dexterous manipulation and multi-robot collaboration. Its embodied reasoning (ER) component, Gemini Robotics ER 2, acts as a high-level brain for planning and human interaction, while an on-device VLA model allows for local operation and rapid adaptation to new robot embodiments.