QCraft unveils Physical AI Model combining World Models and Reinforcement Learning, expands beyond autonomous vehicles into broader real-world AI systems

QCraft announced the QCraft Physical AI Model at Beijing Auto Show, positioning the company's future in Physical AI systems that perceive, reason about, and act in real-world environments, rather than autonomous vehicles alone.

The unified architecture combines World Models (which generate rare and dangerous driving scenarios in simulation using natural language commands) with a Vision-Language-Action model and Reinforcement Learning algorithms, transferring learned capabilities from simulation to production vehicles.

QPilot MAX, a city-level Navigate on Autopilot solution, is deployed across 25 production models from China's largest OEM automotive manufacturer, with 50 additional models expected this year; its Automatic Emergency Braking system has a false activation rate of one per 500,000 kilometers.