[Huatai Securities: focus on the lightweight Application of the AI large Model on the Terminal side] Huatai Securities Research News said that the model compression technology is the core technology to realize the side / end deployment of the AI large model. At present, Google, Microsoft, Tencent and other manufacturers have layout in this field, which will accelerate the integration of AI technology and intelligent terminals, and domestic SoC design companies are expected to benefit deeply. Model compression technology can reduce the demand for reasoning power while keeping the original performance and accuracy of the large model basically unchanged: 1) quantization: transforming floating-point computing into low-bit fixed-point computing; 2) network pruning: removing redundant channels and neuron nodes in the neural network; 3) knowledge distillation: the large model is used as the teacher model and its output is used to train the student model with similar performance and simpler structure.