Google DeepMind

Google DeepMind

"Build AI responsibly to benefit humanity"

Alphabet's London-based AI lab, developer of the Gemini Robotics vision-language-action models and the open-source MuJoCo physics simulator.

Visit Website

Company Overview

Google DeepMind is the artificial intelligence research laboratory of Alphabet Inc., headquartered in London, United Kingdom. DeepMind Technologies was founded in London in 2010, acquired by Google in 2014, and merged with the Google Brain team in 2023 to form Google DeepMind, which develops the Gemini family of models alongside scientific AI systems. In robotics and embodied AI, the lab is one of the most influential developers of robot foundation models and simulation tools.

Its main robotics product line is Gemini Robotics, a family of models that bring Gemini's multimodal understanding into physical action. The first Gemini Robotics models were introduced in 2025. On 30 July 2026 the lab released the second generation. Gemini Robotics 2 is a vision-language-action (VLA) model that converts vision and language input directly into motor commands and can control full humanoids, bi-arm robots, dexterous hands and parallel grippers, including whole-body behaviours such as walking, crouching and reaching, and multi-robot collaboration. Gemini Robotics ER 2 is an embodied reasoning vision-language model that acts as a high-level planner, breaking tasks into steps and coordinating with the VLA, humans and other robots; it is available as a preview in Google AI Studio under the model ID gemini-robotics-er-2-preview and in private preview on Google's enterprise agent platform. Gemini Robotics On-Device 2 is a lightweight VLA optimised to run locally on robot hardware without network connectivity, and DeepMind reports that it can adapt to a new bi-arm embodiment with a few hours of data, typically fewer than 200 demonstrations. The VLA models are offered to early-access partners and more than 100 trusted testers.

Demonstrations in the July 2026 release used the Apptronik Apollo 2 humanoid fitted with SharpaWave 22-degree-of-freedom hands and Inspire hands, a Franka Duo bimanual platform with Robotiq grippers, and Dexmate, SO101 and Trossen platforms for the on-device model. Apptronik, Boston Dynamics and Agile Robots are named as partners.

The second pillar is MuJoCo, a free and open-source physics engine for robotics, biomechanics, graphics and animation that DeepMind maintains and holds the trademark for. MuJoCo is widely used in reinforcement learning and robot learning research, and it is one of the standard simulators for academic manipulation benchmarks. DeepMind also publishes robotics research code and models through its GitHub organisation.

On the hardware side, Google DeepMind appointed former Boston Dynamics chief technology officer Aaron Saunders as vice president of embodied systems and robotics in November 2025. In October 2026 Saunders announced a new robotics hardware lab in Cambridge, Massachusetts, recruiting machinists, engineers and researchers to build physical systems for the lab's AI models. For the RobotToday humanoid and embodied AI landscape, Google DeepMind sits in the foundation models and embodied software segment as a supplier of general-purpose robot models and simulation tools used by humanoid and manipulation robot makers.

Capabilities & Activities

Primary type & automation activities this supplier delivers:

Applications & Industries

Product Categories

Products & Solutions

Gemini Robotics 2

Vision-language-action model converting vision and language into motor control for humanoids, bi-arm robots, dexterous hands and grippers, incl. whole-body control.

Gemini Robotics ER 2

Embodied reasoning VLM that plans multi-step tasks in physical space and coordinates with the VLA, humans and other robots; preview in Google AI Studio.

Gemini Robotics On-Device 2

Lightweight VLA that runs locally on robot hardware without network connectivity and adapts to new bi-arm embodiments from a few hours of data.

MuJoCo

Free, open-source physics engine for robotics, biomechanics and animation, widely used for robot learning and simulation benchmarks; maintained by Google DeepMind.

Partnership & Notable clients

Associating Events

Listed in RobotToday Supplier Discovery. Verified Profile

More related news

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20

Contact Google DeepMind

WEBSITE

https://deepmind.google

HEADQUARTERS

London , England

United Kingdom

Company Facts

Founded

2010

Primary Role

Software/Algorithm

Company Size

-

Primary Region

Europe

Annual Sales

-

Funding Stage

Subsidiary / Acquired

Funding Total

-