Robotics & physical AI Model or AI service
Gemini Robotics 2.5 connects task reasoning and robot control
Google DeepMind introduced Gemini Robotics 2.5 in November 2026. It describes an action model for robot movement working alongside an embodied reasoning model that tracks the steps of a task. [1]
- Model access
- Google DeepMind has published the introduction and demonstration details. [1]
What the model does
DeepMind reported a demonstration in which a robot picked up automotive brackets and placed them into sequencing containers. Its reasoning model tracked workflow while the action model controlled hands and arms. [1]
Why it matters
The example shows how one part of a system can track what remains to do while another handles the physical motion. Each part needs to work for a robot to finish the task. [1]
Key dates
- Google DeepMind introduced Gemini Robotics 2.5 [1]
Ways to use it
Robot developers evaluating the work need to distinguish the reasoning component, the action model and the physical platform used in the demonstration. [1]
Model limitations
The announcement describes company demonstrations. General public checkpoint access and performance in other deployments require separate evidence. [1]
Sources and verification
Catalog evidence checked How we verify records
Related profiles
Robotics & physical AI
NEURA 4NE1
NEURA Robotics
A humanoid platform NEURA presents for development and industrial training, with multiple distinct hardware generations.
Robotics & physical AI
LUMO all-terrain humanoid
Shenzhen Dobot (DOBOT)
LUMO is a humanoid robotics platform whose proposed applications include interaction and service tasks.
Robotics & physical AI
Mobo home companion humanoid
AiMOGA Robotics (Chery Group)
The developer describes Mobo as combining speech and vision for interaction in the home.