Google DeepMind releases Gemini Robotics 2, extending robot control to whole humanoid bodies

Google DeepMind/Gemini Robotics

Google DeepMind released Gemini Robotics 2 on July 30, 2026, a family of three models. Earlier versions mostly handled arms and tabletop tasks; the new vision-language-action model controls a full humanoid, from feet to fingertips. Developers can use the ER 2 reasoning model now, while the control models stay in early access.

• Gemini Robotics 2 is the whole-body motor control model; Gemini Robotics ER 2 plans multi-robot tasks; Gemini Robotics On-Device 2 runs locally on the robot.
• Gemini Robotics ER 2 is available now in Google AI Studio and in private preview on Gemini Enterprise Agent Platform.
• The Gemini Robotics 2 VLA and On-Device 2 are open to early-access partners through a trusted tester form.
• In a demo, Apptronik's Apollo 2 used a 22-degree-of-freedom SharpaWave hand to tie a knot and seal a zip bag.
• The company says the on-device model can adapt to a new robot body within a few hours.