Google unveils Gemini Robotics 2, an AI that coordinates the full body of humanoid robots

A note on the source: the downloaded content of this Bloomberg article (bylined Guinevere Grant and Davey Alba, published on 30 July 2026) is paywalled. What was retrieved contains barely two paragraphs of actual text, surrounded by navigation menus, corporate links and the site's footer. This summary is therefore deliberately brief and confined to what that fragment explicitly confirms, without adding any information that does not appear in the text.

According to the available excerpt, Google DeepMind — Alphabet Inc.'s artificial intelligence division — has unveiled a new AI model for robots called Gemini Robotics 2. It forms part of the company's ongoing effort to extend its Gemini technology to robotics, giving machines the ability to reason, plan multi-step tasks and adapt to environments designed for people.

The development the article highlights is that Gemini Robotics 2 can coordinate movement across a humanoid robot's entire body, from top to bottom, including walking, crouching and manipulating objects while reasoning about the task at hand. The text contrasts this with the company's previous model, which it says controlled mainly the robot's upper body.

The fragment offers no further detail: it names no specific robot manufacturers testing the system, nor any availability dates, prices, performance figures or statements from Google or Google DeepMind executives. Those details would require consulting the full version of the article, which remains behind Bloomberg's paywall.