XPeng Powers on Third Turing Chip, Boosts VLA Response Speed 300%
XPeng announced Aug. 27, 2026 that the third chip in its self-developed Turing AI lineup has been powered on in production vehicles, completing a three-chip architecture for a car-wide super agent. It runs a second-generation VLA merged with a vision-language model, enabling unified scheduling, multimodal understanding, and a larger on-device model. Streaming inference is said to deliver a 300% end-to-end response speed gain, while Infini-VLA adds rolling 30-second road-condition memory.