Gemini Robotics 2 Extends Control From Feet to Fingertips

Google DeepMind released Gemini Robotics 2 in three parts: a vision-language-action model for motor control, an embodied-reasoning model for multi-minute planning, and an on-device variant that runs without network latency. The pitch is whole-body control — locomotion, balance, reaching and manipulation coordinated as one policy rather than a planner bolted to a walking controller. Supported platforms include Apptronik’s Apollo 2, Franka Duo, Dexmate, SO101 and Trossen. Published success rates are wide: 45.7–76.3% on general whole-body tasks and 32–92% on multi-finger work. Only the reasoning model is generally available; the action models are early-access.