DeepMind Tests Robot on Mercedes-Benz Assembly Line

Google DeepMind, Google's AI research lab, released a new robotics model called Gemini Robotics 2.5 and tested it on two Mercedes-Benz car factories, in Hungary and Germany. A humanoid robot called Apollo 2, built by a company named Apptronik, used the model to sort parts from bins into the order each car on the line needed. DeepMind said the robot completed the task with 96% accuracy. Compared with the previous version of the model, DeepMind reported a 30% to 100% higher success rate, depending on the task, and work done 20% to 30% faster.

The test ran on a working assembly line, not in a lab, and production lines are not stopped for experiments. That raises the weight of these numbers. If a robot can hit that accuracy while working near people on a line, carmakers have a reason to hand it routine parts-sorting without slowing down assembly. For the market for humanoid robots, a result from an ordinary factory floor carries more weight than a trade-show demonstration.