DeepMind introduced Gemini Robotics 2 on July 30, 2026, a new AI model designed to enable robots with whole-body intelligence. The system teaches robots fine dexterity, intelligent control from feet to fingertips, and teamwork capabilities to perform complex, varied tasks. This marks a step toward robots that can adapt and learn autonomously beyond pre-programmed routines, according to deepmind.google.
Gemini Robotics 2 builds on previous robotic AI by allowing skill transfer across different robot bodies and improving adaptability in unpredictable environments. Unlike traditional robots limited to narrow, repetitive tasks, this model enables robots to learn and generalize skills. DeepMind demonstrated the system’s ability to coordinate multiple limbs and collaborate on tasks, highlighting advances in fine motor control and multi-agent teamwork.
The development addresses longstanding challenges in robotics, such as the inability to generalize learned skills and the reliance on teleoperation or rigid programming. Gemini Robotics 2’s whole-body intelligence could accelerate applications in manufacturing, logistics, and service robots. This advancement aligns with broader AI trends emphasizing embodied intelligence and multi-modal learning, positioning DeepMind alongside leaders in robotic AI research.
DeepMind published the Gemini Robotics 2 announcement on its official blog on July 30, 2026, including demonstration videos showcasing the model’s capabilities in real-world robotic tasks.