Get the best builds by email. No account needed.
Live drop
Google DeepMindDropped Nov 10, 2026
Higher dexterity and 30-100% better success on 200+ real-world robot tasks with whole-body control.
About this release
Gemini Robotics 2.5 is Google DeepMind's latest vision-language-action (VLA) model for controlling robots of any embodiment. It improves task breadth, learning speed, dexterity (including 22-DoF hands and deformable objects), reliability, and physical safety steerability over Gemini Robotics 2. It achieved 96% success in a Mercedes-Benz factory pick-and-pack proof-of-concept with Apptronik's Apollo 2 humanoid and rolls out to robotics partners.
Read the official Google DeepMind launch pageFirst builds incoming — this lane fills in automatically as people start building with Gemini Robotics 2.5. Know one? Submit the URL.