XPeng Motors officially announced that its third Turing chip has been powered on! With the super intelligent agent on board, a single sentence can drive the car.

XPeng announced on Weibo today that its third Turing AI chip has been officially lit up, bringing the super intelligent agent onboard. The XPeng Physical AI Sharing and Second-Generation VLA New Version Experience Day will be held at 15:00 today.

According to XPeng, with the third Turing chip lit up, XPeng's second-generation VLA and VLM achieve cross-domain integration, featuring cross-domain total scheduling, native multimodality, and on-device large model capabilities.

The in-cabin voice architecture has been fully restructured, supporting voice-activated pull-over parking, nearby parking spot selection, fuzzy navigation, and complex trip planning. It can "listen and see, reason well, and express itself," with one brain managing the entire vehicle and a single command controlling the car.

The Turing AI chip is XPeng's self-developed, world's first multi-terminal general-purpose AI chip, applicable to AI cars, flying cars, and humanoid robots. It has secured mass-production designation from Volkswagen and launched external commercial collaborations.

Previously, models like the XPeng G7 and the all-new P7 were equipped with the Turing AI chip to drive the VLA intelligent driving large model, with total effective computing power reaching up to 2250 TOPS. With the third chip lit up, it drives the cross-domain integration of VLA and VLM, evolving the vehicle from a mere intelligent driving tool into a "super intelligent agent."

It is understood that XPeng's second-generation VLA was officially released and began rollout in March this year. It is a physical world large model aimed at fully autonomous driving, integrating "seeing, hearing, and reading," supporting P-gear in-place activation of full-scenario assisted driving, covering scenarios such as highways, narrow roads, and map-free parking lots.

During testing, it also demonstrated emergent intelligent capabilities such as recognizing traffic police gestures and proactively responding to traffic light transitions, which were not explicitly trained.

This new version connects the VLA intelligent driving system with the VLM multimodal large model, meaning the vehicle can not only drive autonomously but also understand fuzzy voice commands and execute complex operations. For example, when a user says "pull over," the vehicle can automatically plan a route and pull over.

With the official debut of XPeng's second-generation VLA new version, XPeng's assisted driving capabilities have been further upgraded. He Xiaopeng previously stated that fully autonomous driving will fully arrive within the next 1 to 3 years, and autonomous driving will truly become people's daily travel habit.