Google DeepMind introduced Gemini Robotics 2.5 on November 10, 2026, calling it its most capable vision-language-action model ...