On November 10, Google DeepMind introduced Gemini Robotics 2.5 and began rolling it out to its robotics partners. The model turns visual input and language instructions into physical actions. Google says its tests across more than 200 distinct tasks showed gains in dexterity, including handling flexible objects and using everyday tools.
One proof of concept took place on Mercedes-Benz production lines in Hungary and Germany. Apptronik’s Apollo 2 humanoid sorted metal brackets into containers in the order they would be needed. Google reports a 96% success rate for the pilot. Gemini Robotics ER kept track of the workflow while Gemini Robotics 2.5 controlled the robot’s hands and arms, including adjustments when gaps appeared.
