LaunchJul 30, 2026

Google DeepMind unveils Gemini Robotics 2 with whole-body control

What's the deal?

Google DeepMind on July 30, 2026, introduced Gemini Robotics 2, an intelligence layer for robots that adds whole-body control, advanced dexterity, and multi-robot collaboration. It is a vision-language-action (VLA) model that converts vision and language input into motor control.

What's the endgame?

The company wants robots that can think, act, and adapt across different bodies. Gemini Robotics 2 lets a humanoid walk, crouch, stretch, and manipulate objects to clean a cluttered room — and team up with other robots to finish faster.

Why now?

Most robots today are pre-programmed or teleoperated for narrow, repetitive tasks, DeepMind says. They struggle to learn on their own, adapt to unpredictable environments, or transfer skills from one robot body to another.

The model builds on Gemini Robotics, which DeepMind used to show how Gemini's multimodal understanding could drive real-world action. The new version can run locally on-device and adapt to entirely new robotic bodies in a few hours.

The signal:

DeepMind is pushing to make a single AI model the standard brain for robots of every shape and size. If skills transfer across bodies as claimed, it lowers a long-standing barrier to deploying robots at scale.

In this story

GGoogle DeepMindDDemis Hassabis

Go deeper

Analysis

The largest AI companies by headcount

Frontier AI labs and data-AI platforms ranked by current team size.

Analysis

The race to put a reactor on a ship

A market map of the nuclear-reactor developers building toward ship propulsion. X = years each has spent developing its reactor (= company age for the nuclear-native firms); Y = total funding raised (Dealroom, log). Colo

Analysis