Google DeepMind just dropped Gemini Robotics ER 2, and it’s the kind of update that makes you rethink what robots are actually capable of. The new model doesn’t just make individual robots smarter. It makes groups of completely different robots work together like they share the same brain.
Boston Dynamics’ Spot and Apptronik’s Apollo, two robots that look nothing alike and do very different things, can now coordinate through a shared semantic understanding powered by Google’s AI.
What Gemini Robotics ER 2 actually does
The core innovation here is what DeepMind calls “temporal intelligence.” In English: the robot can watch what’s happening through a continuous video feed, understand where it is in a multi-step task, and plan its next move accordingly.
The numbers tell an interesting story. The model hits 57.4% accuracy for progress classification and 91.3% for moment finding, with a mean error of just 0.96 seconds. Progress classification, the ability to understand how far along a task is, still has room to grow. But moment finding, pinpointing exactly when a specific event occurs in a video stream, is remarkably precise at sub-second accuracy.
