Uploaded September 2025 | Updated September 2026, 1 week ago
Robots come in all shapes and sizes, and have different sensing capabilities and different degrees of freedom, making it difficult to transfer motions learned from one robot to another.
Gemini Robotics 1.5 shows a remarkable ability to learn across different embodiments. It can transfer motions learned from one robot to another, without needing to specialize the model to each new embodiment. This breakthrough accelerates learning new behaviors, helping robots become smarter and more useful.
Learn more at https://deepmind.google/models/gemini-robotics/
___
Subscribe to our channel youtube.com/@googledeepmind
Find us on X twitter.com/GoogleDeepMind
Follow us on Instagram instagram.com/googledeepmind
Add us on Linkedin linkedin.com/company/deepmind
Robots come in all shapes and sizes, and have different sensing capabilities and different degrees of freedom, making it difficult to transfer motions learned from one robot to another.
Gemini Robotics 1.5 shows a remarkable ability to learn across different embodiments. It can transfer motions learned from one robot to another, without needing to specialize the model to each new embodiment. This breakthrough accelerates learning new behaviors, helping robots become smarter and more useful.
Learn more at https://deepmind.google/models/gemini-robotics/
___
Subscribe to our channel youtube.com/@googledeepmind
Find us on X twitter.com/GoogleDeepMind
Follow us on Instagram instagram.com/googledeepmind
Add us on Linkedin linkedin.com/company/deepmind










