Gemini Robotics 1.5: Learning across embodiments @googledeepmind
Gemini Robotics 1.5: Learning across embodiments  @googledeepmind
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
Gemini Robotics 1.5: Learning across embodimentsIntelligent whole-body control with Gemini Robotics 2SynthID: A tool for watermarking and identifying AI-generated contentThis is Gemini Omni ♾️ #GoogleIOGemini Robotics 1.5: Using agentic capabilitiesVeo 3.1 - Create longer, seamless shotsGemini 3: Bring anything to lifeVeo 3.1 and more artistic control in FlowVeo 3 demo | Chrome cityProject Genie Gallery | Cat VacMulti-robot collaboration with Gemini Robotics 2Using AI to outsmart drug-resistant bacteria
Google DeepMind |

Gemini Robotics 1.5: Learning across embodiments

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER