On February 12, Xiaomi unveiled that it would make its first-generation robotics VLA model, Xiaomi-Robotics-0, open-source. This model is impressive, featuring a staggering 4.7 billion parameters. It seamlessly integrates visual language comprehension with high-performance real-time operational capabilities. It has achieved numerous SOTA (State-of-the-Art) benchmarks in both simulated tests and real-world robotic tasks. Additionally, it can perform real-time inference using consumer-grade graphics cards.
