Meituan Accelerates Vision AI Inference Services and Optimizes Costs
Meituan Accelerates Vision AI Inference Services and Optimizes Costs
1 PagesMeituan leverages vision AI to drive business innovation and deliver personalized services but faced challenges in balancing performance and cost. To optimize AI inference, especially for low-traffic, long-tail models, Meituan turned to 4th Gen Intel® Xeon® Scalable processors with Intel® Advanced Matrix Extensions (AMX). This CPU-based solution, combined with dynamic scaling and service optimization, enabled high-throughput inference without sacrificing accuracy. The result was a 4.13x boost in model performance using BF16 and up to a 3x increase in online resource efficiency, while reducing service costs by 70%.
