Xpeng launches GX robotaxi closed beta testing

Xpeng has officially started closed beta testing of its robotaxi program; Chairman and Chief Executive Officer He Xiaopeng became the first passenger to order a ride from outside the company’s Guangzhou headquarters, with what he described as a completely seamless greeting process. This milestone spans eight months, from announcing the robotaxi plan last November, to road testing in January, to production in May and now an end-to-end operating service. The vehicle is built on Xpeng’s newly launched flagship GX SUV and represents China’s first fully in-house developed, factory-assembled, mass-produced robotaxi. The vehicle is designed to SAE Level 4 autonomous driving standards and runs on four of Xpeng’s own Turing AI chips, providing a total of 3,000 TOPS of onboard computing power. Clearly aiming to future-proof the vehicle as technology improves, Xpeng claims this is the highest level of any vehicle currently available globally. Unlike Waymo’s LiDAR and HD mapping-heavy approach, Xpeng has built its robotaxi as a camera vision-driven production vehicle from day one, treating autonomy primarily as a manufacturing issue rather than a fleet conversion exercise. The company’s approach is similar to Tesla’s in terms of which sensors it prioritizes, but unlike Tesla it has a radar system. The VLA 2.0 model handles driving decisions end-to-end, reducing system latency to less than 80 milliseconds and allowing the car to be generalized to new cities without pre-mapped data; a real point of differentiation, considering capital-intensive HD mapping and how restrictive geofencing has proven for competitors. He Xiaopeng, Chief Executive Officer at headquarters, was the first person to take Xpeng’s robotaxi beta for a test drive. This manufacturing-first logic also extends to Xpeng’s business structure. A dedicated Robotaxi Business Unit, established in March, directly oversees the fleet, while a parallel SDK strategy allows third parties such as Amap, which has already signed on as the first ecosystem partner, to dispatch Xpeng robotaxis through in-house applications rather than requiring Xpeng to create a ride-hailing request from scratch. The company is also planning a “Robo” hardware that would apply the same autonomous technology used in its commercial fleet to cars sold directly to regular consumers. It’s a route neither Waymo nor Tesla’s custom Cybercab have followed in the same way. Because Robotaxi shares its GX underpinnings with a mass-market consumer vehicle, Xpeng can also leverage its existing dealer service and charging network to maintain the fleet, avoiding dedicated warehouse costs that burden software-only operators. The VLA 2.0 model is not only aimed at self-driving functions: it also serves as the joint basis for Xpeng’s Iron humanoid robot and the Aridge (formerly AeroHT) flying car program. Together, these programs reflect a broader claim that training the same AI across cars, robots, and airplanes will mature driving intelligence faster than traditional simulation alone. Pilot operations are planned for the second half of 2026, and the company is targeting fully driverless service (without safety operators) in early 2027. In a social media post, Xpeng’s He was careful to frame the closed beta milestone as an interim first step towards deployment, saying large-scale rollout was still a long-term process dependent on security, experience and generalization gains. The GX itself went on sale in China in May for CN¥269,800 (US$39,700); This means that Xpeng is currently testing its autonomy stack on a vehicle that is already generating traditional retail revenue; this is a measure that limits the commercial risk of the robotaxi program running behind schedule.