China has a large population, and the demand for agricultural products and food is great and diverse. Moreover, the quality and safety of agricultural products and foods are closely related to people's daily life. Therefore, it is the development needs of contemporary society to use nondestructive, rapid, environmentally friendly and high-through put testing methods to detect the quality of agricultural products and foods. The traditional detection and analysis methods have some disadvantages, such as time and labor consumption, the tested samples cannot be sold again after testing and the defective products missing detection. As a rapid and nondestructive detection method, near-infrared spectroscopy (NIR) has been paid more and more attention by some scholars and related industry personnel. However, most NIR analysis methods only build mathematical models for single material. For a large number and variety of agricultural products and foods, such as different regions, different years, different temperatures, different processing methods, different components and even different varieties, this relatively traditional NIR analysis method will undoubtedly increase the workload of preliminary modeling. With the development of computer technology, spectrometer hardware, stoichiometry and internet technology, relevant scholars have begun researching and developing a universal NIR model to solve this problem. That is to establish a near-infrared universal model, which can detect the same index or multiple indexes of various materials. Compared with the traditional NIR model, the universal model has the advantages of low modeling cost and small workload, which makes the application and promotion of NIR spectroscopy technology in agricultural products and foods field of great significance. This paper reviews the research on the universal model of NIR in detecting agricultural products and foods. By comparing the traditional model modeling method with the universal model modeling method, the methods used in the three modeling steps of sample information acquisition, model establishment and sample information prediction in building the universal model are summarized. At the same time, the main points of modeling of NIR universal model in agricultural products and foods detection are summarized. Currently, the research of the NIR universal model in agricultural products and food quality detection is still in the development stage. In this paper, some suggestions on the development and research of the universal model are proposed, and the development trend of the NIR universal model in agricultural products and foods detection has further prospected.