# Gemini Robotics 2 moves Google’s robot model toward whole-body control

Canonical URL: https://www.theroboteconomy.news/articles/gemini-robotics-2-whole-body-control/

Publisher: The Robot Economy

Editorial byline: Imani Reed, Frontier Desk

Published: 2026-07-30T20:10:48.491Z

Updated: 2026-07-30T20:10:48.491Z

Editorial information: https://www.theroboteconomy.news/about/#editorial-desks

Google’s July 30 model release adds whole-body control, dexterity and multi-robot planning, but the motor-control model remains private preview.

## Summary

Google DeepMind’s Gemini Robotics 2 release expands its robotics model family from upper-body task control toward whole-body humanoid control, dexterous manipulation, multi-robot planning and faster on-device adaptation. The evidence is a vendor announcement and early-access documentation, not customer deployment or independent field evaluation.

## Reader context

Vision-language-action model: A model that turns visual and language input into physical robot actions.

Embodied reasoning: Planning that uses a robot’s physical surroundings and delegates actions to a lower-level controller.

## Key facts

- Google DeepMind introduced three Gemini Robotics 2 models for motor control, embodied reasoning and on-device robot control.

- The company shows one model checkpoint across Apollo 2 and Franka Duo configurations and claims whole-body humanoid control.

- Google lists Gemini Robotics 2 as private preview, while Gemini Robotics ER 2 is available through the Gemini API and Google AI Studio.

- The published evaluation includes a 45.7 percent whole-body task result and several five-finger tasks below 50 percent success.

## Lead image

Image: https://www.theroboteconomy.news/images/articles/gemini-robotics-2-whole-body-control.png

Description: A generic humanoid robot and compact wheeled robot coordinate around a simple object in a conceptual workspace.

Credit: AI-generated editorial illustration / The Robot Economy

## Report

### What changed

Google DeepMind introduced Gemini Robotics 2 on July 30 as a family of three models: a vision-language-action model for motor control, an embodied-reasoning model for planning and multi-robot coordination, and an on-device model for local inference. The release claims a shift from upper-body, tabletop tasks toward whole-body control: its Apollo 2 example walks, bends and places a watering can, while the same checkpoint is shown across Apollo 2 and Franka Duo configurations.

Sources: [s1](<https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/>) [s2](<https://deepmind.google/models/gemini-robotics/vla/>)

The decision delta is not that robots are now general-purpose operators. It is that the model layer is being packaged around body transfer and orchestration: ER 2 can watch video, track task progress and hand off actions, while On-Device 2 is said to adapt to new embodiments in a few hours with fewer than 200 examples. That moves the integration question from single-arm task learning toward coordination across bodies and time.

Sources: [s1](<https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/>) [s3](<https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/>)

### The boundary

Availability remains the limiting fact. Google lists Gemini Robotics 2 as private preview and directs developers to an early-access partner program; ER 2 is public through the Gemini API and Google AI Studio, with private preview on the enterprise agent platform. Google’s published figures include 45.7 percent on one whole-body task and 32 to 44 percent on several five-finger tasks, even as the company describes the release as a step toward general-purpose physical AI.

Sources: [s1](<https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/>) [s2](<https://deepmind.google/models/gemini-robotics/vla/>) [s3](<https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/>)

Axios independently reported the release as an AI-software update, not a customer deployment. No source opened here establishes production use, independent evaluation or field reliability. The next checkpoint is an early-access partner or independent lab publishing repeatable results on a new embodiment with disclosed intervention rules, task duration and failure rates.

Sources: [s1](<https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/>) [s2](<https://deepmind.google/models/gemini-robotics/vla/>) [s3](<https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/>) [s4](<https://www.axios.com/2026/07/30/google-robotics-software-update>)

## Conclusion

The article compares Gemini Robotics 2’s whole-body, multi-robot capability claims with the access and evaluation boundary that actually governs near-term use: private preview for motor control, public access for the reasoning layer, vendor-led tests and no customer operation.

## Disclosures

Sponsored: no

Desk bylines, AI assistance and editorial policies: https://www.theroboteconomy.news/about/

## Sources

- s1: [Gemini Robotics 2 brings whole body intelligence to robots](<https://deepmind.google/blog/gemini-robotics-2-brings-whole-body-intelligence-to-robots/>) — Google DeepMind; company-announcement; 2026-07-30T00:00:00.000Z

- s2: [Gemini Robotics 2](<https://deepmind.google/models/gemini-robotics/vla/>) — Google DeepMind; product-documentation; 2026-07-30T00:00:00.000Z

- s3: [Introducing Gemini Robotics ER 2](<https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/>) — Google Developers Blog; company-announcement; 2026-07-30T00:00:00.000Z

- s4: [Google updates its AI software for robots](<https://www.axios.com/2026/07/30/google-robotics-software-update>) — Axios; independent-reporting; 2026-07-30T15:00:06.000Z