Tom Power

Robotics at @GoogleDeepmind

[Tom Power @thomas_j_power]

Excited to share what we've been working on! 🤖

With Gemini Robotics 2 we can bridge the sim2real gap and learn new tasks entirely from simulation!

Here it is performing an automotive kitting task, handling bimanual grasping, 3D reorientations, and tight insertions. 👇

[Todor Davchev @tdavchev]

Gemini Robotics 2 can also generalise to a range of scenes including out of distribution ones while maintaining it's precision. Notice how it also works across different heights and 3D poses.

[Maria Bauza Villalonga @bauzavillalonga]

After years of working on precise dexterous manipulation, I am thrilled to share our latest milestone in robotic dexterity! 🤖🗑️

Gemini Robotics 2 achieves advanced spatial awareness, precise control & robust VLA capabilities in multi-finger hand control.
Great team success🦾💡

[Norman Di Palo @normandipalo]

We got many comments saying the release videos are cool, but do the models actually work?

We stress tested Gemini Robotics 2 to show its impressive robustness and recovery capabilities.

It's amazing to see how it keeps trying until the task is done

[Tom Power @thomas_j_power]

Really happy to see this shared!

Gemini Robotics 2 performing tasks requiring 3D reorientations and precision insertions uninterrupted for 20 minutes, all while recovering from failures and disturbances !

[Konstantinos Bousmalis @bousmalis]

Here’s a first look at Gemini Robotics 2 from @GoogleDeepMind on the FR3 Duo: 20 minutes of uninterrupted, real-time tool kitting.

Notice the emergent recovery behaviours - generalized dexterity & precision at a whole new level.

More demos dropping over the next days! Stay tuned!

[Google DeepMind @GoogleDeepMind]

One brain. For any robot. 🤖

We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.