Google DeepMind Announces Gemini Robotics ER 2, an Embodied Reasoning Model for Robots

Google DeepMind announced Gemini Robotics ER 2, an embodied reasoning model for robots, on July 30, 2026. The model serves as a high-level brain for robots, handling dialogue, understanding the physical world and planning multistep tasks, while leaving motor control to lower-level VLA models. An upgrade from the previous-generation ER 1.6, it newly introduces self-progress tracking through continuous video feeds and multirobot coordination. It will be available to developers through the Gemini API, Google AI Studio and Gemini Enterprise Agent Platform (private preview).

For developers building applications for robots, the model is important as a higher-level system that handles dialogue, understanding the physical world and planning multistep tasks, while handing control to lower-level vision-language-action models. It can track progress through continuous video, retry after failures or switch to the next step, making it relevant to managing complex workflows.