Gemini Robotics 2 brings whole body intelligence to robots
Google's Gemini Robotics 2 unifies vision and motor control in a single model—the first to run end-to-end on real hardware.
Why it matters
A capability breakthrough in embodied AI: multimodal models that reason about the physical world and execute complex motor tasks in real time. This changes what's possible in robotics and signals Google's competitive push against OpenAI's physical AI work.
The key facts
6 to knowGemini Robotics 2 announced by DeepMind/Google
End-to-end vision-to-motor model (whole-body control in one neural net)
Deployed and tested on real robotic hardware
Multimodal reasoning: visual perception + motor planning unified
Positions Google in the embodied AI race against OpenAI's physical AI focus
Published July 28, 2026 on DeepMind blog
Go to the source
Google DeepMind Blogdeepmind.google