At today's XPENG 2025 Tech Day event, CEO He Xiaopeng unveiled the company's second-generation VLA (Vehicle Learning Architecture), a groundbreaking physical model framework applicable across AI-powered vehicles, Robotaxis, robotics, and flying cars.
He highlighted the architecture's exceptional autonomous driving capabilities, demonstrated by features like seamless NGP (Navigation Guided Pilot) transitions across highways, urban roads, and narrow pathways. In rigorous 20km complex road tests, the second-gen VLA required only one manual intervention—a 13x improvement over previous systems—while Tesla's FSD needed five interventions.
The system also exhibits emergent intelligence, including real-world contextual understanding such as interpreting traffic police gestures and anticipating traffic light patterns.
Notably, XPENG will debut its 'Super LCC+' navigation-free autonomous driving assistant, compliant with global regulations and operable anywhere.
"Select users will join a beta program by late December, with full rollout to XPENG Ultra models slated for Q1 2026," He announced.