联想发布端侧AI新技术，让PC本地推理能力媲美 OpenAI-mini

Core Viewpoint - Lenovo showcased the "Lenovo Inference Acceleration Engine" at the Tech World conference, designed for efficient AI PC inference, achieving local inference capabilities comparable to OpenAI's o1-mini cloud model released last year [1] Group 1: Product Features - The inference engine is a software-hardware collaborative optimization platform, enhancing the local inference capabilities of a standard PC [1] - It utilizes high-performance parallel decoding, operator fusion optimization, and heterogeneous computing technology to significantly improve inference speed across various application scenarios [1] - The engine also reduces power consumption and memory usage, contributing to overall efficiency [1] Group 2: Performance Enhancement - When combined with large model inference chips, the engine leverages seamless collaboration between hardware acceleration and software optimization, resulting in compounded performance improvements [1]