XPeng Motors recently unveiled its second-generation VLA (Vehicle Language Assistant) intelligent driving technology, which emphasizes a "veteran driver-like" natural driving experience and specifically addresses three major pain points users face with current intelligent driving systems: abrupt maneuvers, poor handling of complex scenarios, and low traffic efficiency. Built upon a native multimodal large model of the physical world, the system marks a leap from "rule-based execution" to "commonsense reasoning," significantly enhancing driving smoothness and safety—reducing hard braking by 99%, sudden acceleration by 98%, jerky movements by 89%, and safety-related driver takeovers by 60%. The new system operates without reliance on high-definition maps or clearly marked lane lines, covering everything from major roads to intricate micro-level scenarios. Starting late March, the technology will be rolled out first on models including the P7 Ultra, G7 Ultra EV, and X9 Ultra, expanding in April to Ultra and Ultra SE variants of the 2026 P7+, G9, and G6. Underpinning this advancement are XPeng’s self-developed Turing chip, 50 petabytes of real-world production data, and over 500,000 simulated driving scenarios, propelling intelligent driving into a new era of "AI instinct-driven" autonomy.