Intel / article

Gemini Robotics 2

Source: deepmind.google
Date: Published Jul 28, 2026

Context

Google DeepMind's Gemini Robotics 2 separates planning a physical task from producing the movements needed to perform it. Its announcement describes an embodied- that interprets instructions, observes progress, and coordinates with a vision-language-action model that turns visual and language inputs into motor control.

The company says the action model extends control to whole-body humanoid movements, while a separate on-device version supports local execution. This division is intended to connect longer task plans with physical actions and corrections. According to the announcement, the reasoning model is available through Google AI Studio and a private preview, while the action and on-device models are limited to early-access partners; it provides no evidence of general reliability in ordinary homes.

Terms in this piece · Glossary

reasoning model — A model trained to think — generating extended internal reasoning before answering — trading time and tokens for accuracy on hard problems.

Read the source deepmind.google