Introducing Gemini Robotics ER 2

Google has announced the release of Gemini Robotics ER 2, its most capable embodied reasoning model designed to serve as a high-level brain for robots. The new model enhances physical agent capabilities by enabling real-time spatial reasoning, tool orchestration, and multi-step task planning while simultaneously executing actions. Key upgrades include advanced temporal intelligence for progress tracking, precise moment-finding, and support for multi-robot collaboration across diverse platforms. Additionally, the model incorporates improved spatial intelligence and robust safety features, such as human proximity detection and safety instruction following. Gemini Robotics ER 2 is now publicly available to developers through the Gemini API and Google AI Studio.

(Source：Gemini)