Google DeepMind Releases Robot Model Gemini Robotics 2.5, Achieves 96% Success Rate in Mercedes-Benz Production Line Pilot

November 11: According to a Google DeepMind announcement, Google DeepMind has released the vision-language-action (VLA) model Gemini Robotics 2.5, which it says is its most capable robot model to date. Compared with previous baseline models, the new model improves task success rates by 30% to 100% and completes tasks 20% to 30% faster, with testing covering more than 200 tasks. In a pilot with Apptronik and Mercedes-Benz, the Apollo 2 humanoid robot autonomously sorted components on automotive production lines in Germany and Hungary, achieving a 96% success rate. The model is being progressively opened to robotics partners.