Li Auto Rolls Out Mach VLA 2.0 Autonomous Driving Model to 1 Million Vehicles
Li Auto has officially begun rolling out its OTA 8.6 software update, bringing the self-developed Mach VLA 2.0 autonomous driving model to nearly one million AD Max vehicle owners. The update spans various models from 2022 to 2025 running on NVIDIA Orin and Thor hardware platforms following months of model distillation and validation. This update represents one of the largest real-world deployments of a Vision-Language-Action (VLA) model in consumer electric vehicles. It demonstrates how automakers can deliver cutting-edge AI capabilities to older vehicle generations without leaving existing customers behind. The Mach VLA 2.0 update applies knowledge distillation to optimize perception, reasoning, and planning tasks across different vehicle hardware variants. Additionally, the OTA 8.6 release introduces voice-command parking integration with Li Auto's Livis AI glasses alongside new display features.
## BACKGROUND
Vision-Language-Action (VLA) models combine visual perception, natural language reasoning, and motor control into a unified policy for autonomous driving tasks. To execute these computationally heavy models on vehicle chips, developers rely on model distillation, a technique that transfers knowledge from a large teacher model into a smaller, highly efficient student model suited for edge devices.