Li Auto Unveils In-House Chip Mach M100 with 1280 TOPS, Set for L9 Livis

On May 12, Li Auto unveiled its self-developed chip, the Mach M100, fabricated using a 5nm automotive-grade process and delivering a single-chip computing power of 1,280 TOPS. Featuring the industry’s first edge-side dynamic dataflow architecture, the chip is designed to meet the high-concurrency and high-compute demands of the AI era. Compared to traditional GPU architectures, it significantly reduces latency and power consumption while boosting computational performance. Through co-development of models and chips, end-to-end latency has been reduced by 40%, enabling vehicle response speeds twice as fast as human reaction times. The Mach M100 will debut in the all-new Li Auto L9 Livis model, serving as the core of its intelligent mobility solution. This solution includes the 3DViT perception model, which fuses LiDAR and visual data to extend visible range by 50% and accurately reconstruct semantic details of the real world. Additionally, the upgraded Mach VLA 2.1 system delivers a tenfold increase in multimodal computation, supporting advanced features such as autonomous driving, automated parking, and intelligent interaction, thereby enhancing capabilities in risk prediction and intent understanding. Li Xiang, founder of Li Auto, stated that the team decided four years ago to adopt a dynamic dataflow architecture and established a foundational technology team for long-term investment. The launch of the Mach M100 validates the feasibility of this technical approach.