Gemini Robotics 2: A New Generation VLA System for Humanoids

From the soles of the feet to the fingertips—we are giving robots intelligent full-body control, fine dexterous manipulation, and collaborative operation capabilities so they can complete a wide range of complex tasks.

For decades, we have envisioned robots seamlessly integrated into human life, ready to lend a hand at any time. Now, this vision has taken a key step forward.

The vast majority of robots rely on preprogramming or teleoperation to execute single, repetitive task flows; they cannot learn autonomously and struggle to adapt to environments full of unknown changes. In addition, transferring learned skills from one robot body to another is extremely difficult. To tackle various highly challenging operational scenarios at scale, robots of all forms and sizes need accompanying AI models that give them the intelligence to think, act, and interact autonomously, and to complete various tasks safely.

We have previously shown that Gemini's multimodal understanding capabilities can drive physical devices to perform real-world operations, through the corresponding technical solution Gemini Robotics.

Today, we are officially launching Gemini Robotics 2—the intelligent foundation supporting the next generation of truly adaptive robots. With this solution now available, this major technological breakthrough delivers three core capabilities: intelligent full-body motion control, advanced dexterous manipulation, and multi-robot collaborative operation.