The embodiment of the invention provides a model loading method applied to a server, and the method comprises the steps: carrying out the layering of a model, obtaining layered models with different precisions, and enabling each layered model to be composed of triangular patches with the corresponding precision; performing double marking on vertex data of all triangular patches according to the precision of the hierarchical model from coarse to fine; wherein the first mark information is the number of a hierarchical model with the coarseness of the vertex data, and the second mark informationis the serial number of the vertex data in a preset vertex sequence; receiving a download request sent by the terminal, recording loading condition parameters of the terminal in the request, selectingvertex data of the target hierarchical model according to the loading condition parameters, and returning the vertex data to the terminal for the terminal to load; wherein the precision of the targethierarchical model is in positive correlation with the loading condition parameters. The embodiment of the invention solves the problem that a terminal with relatively low performance and relativelypoor network environment may have a jamming phenomenon when loading and rendering a model with relatively high fineness in a traditional scheme.