Daily Car
·01/05/2026
Xpeng's new VLA 2.0 advanced driver assistance system is now being deployed, presenting a notable challenge to established systems like Tesla's Full Self-Driving (FSD). Initial real-world tests reveal a system with a distinct operational feel and a different underlying technology.
Instead of relying primarily on data collected from cars on the road, as Tesla does, Xpeng uses what it calls a "generative world model." This is a powerful computer simulation that can create and test millions of driving scenarios virtually. This method allows Xpeng to simulate the equivalent of 30 million kilometers of driving every day. Like recent versions of Tesla's system, VLA 2.0 primarily uses cameras for its vision, a shift away from earlier systems that also used lidar, a sensor that uses laser light to measure distances.
During recent demonstrations in Beijing's complex traffic, the VLA 2.0 system operated with notable smoothness. It navigated dense traffic confidently, changing lanes and maneuvering through tight spaces in a manner consistent with local driving styles. The system's assertiveness allowed it to blend in with human drivers, rather than appearing overly cautious or robotic in its decision-making.
While the performance appears polished, the system is not without areas for improvement. Observations from test rides indicate it can sometimes follow the vehicle ahead too closely. At other times, it showed hesitation in preparing for upcoming turns or moving into the correct lane in a timely manner. When compared to the version of Tesla's FSD available in China, VLA 2.0 was reported to feel smoother, though the full scope of its emergency handling capabilities has not yet been demonstrated.
The arrival of VLA 2.0 highlights the rapid pace of technological development in China's automotive sector. While Tesla has long been a benchmark for driver assistance technology, competitors are quickly closing the gap with different approaches. The competition is intensifying, suggesting that the leadership position in this field will face significant challenges from new and capable systems in the near future.