FrontierThe story, in brief

Google DeepMind’s new AI model can control a robot’s entire body

Google DeepMind's Gemini Robotics 2 now controls a humanoid robot's entire body—from feet to fingertips. That changes what robots can actually do.

Illustration of a transparent lens revealing connected networks across layers of paper.
Exploring the next frontier of AI research.AI illustration by KeyNews
The KeyNews take

Why it matters

A capability jump in robotic control: the model progressed from upper-body manipulation to whole-body coordination (walking, crouching, object manipulation). This is a frontier model advancement with immediate hardware implications—practitioners evaluating robotics AI stacks need to track this.

The key facts

5 to know
  1. Gemini Robotics 2 supports whole-body motions (previous version: upper body only)

  2. New capabilities: walking, crouching, stretching, object manipulation

  3. Demonstrated on Apptronik Apollo 2 robot

  4. Specific tasks shown: bending to pick up objects, shelf retrieval

  5. Published by Google DeepMind, July 2026

Go to the source

The Verge AItheverge.com

Publisher excerpt: Apptronik’s Apollo 2 robot takes a baseball glove off of a shelf. | Image: Google Google DeepMind says the latest version of its Gemini Robotics AI model can "control entire humanoid robots." While the previous model focused on controlling a humanoid robot's upper body, Gemini Robotics 2 now…
Read original report
Back to today's editionMore frontier news

Keep reading

Related stories

More from Frontier