DeepRoute.ai Presents Physical AI Vision and Foundation Model at Auto China 2026

DeepRoute.ai presented its latest progress in Physical AI at the 19th Beijing International Automotive Exhibition, announced in a press release. CEO Maxwell Zhou outlined the company's goal to build AI infrastructure for the physical world, while Chief Scientist Chong Ruan introduced updates to the company's Foundation Model for advanced intelligent driving.

Zhou stated that current intelligent driving systems have achieved safety levels several times higher than human drivers, though improvement is still required. He said DeepRoute.ai aims to deliver fully safe advanced driving systems within two to three years as large models improve their understanding capabilities.

Ruan described the Foundation Model as a unified architecture that integrates driving decisions, scene comprehension, and behavior evaluation. The model uses high-quality data and a faster data-driven closed-loop, reducing the iteration cycle from about five days to twelve hours. This approach is designed to improve stability, safety, and efficiency in large-scale deployment.

DeepRoute.ai also previewed its Cabin-Driving Integration Agent, which functions as an AI Brain to interpret user intent and respond to complex driving conditions. The company reported more than 300,000 vehicles equipped with its Urban NOA solution and over 1.3 billion kilometers of road data collected. By 2026, it plans to exceed one million mass production deliveries and increase its Miles Per Critical Intervention metric to over 1,000 kilometers.