Abstract: The debris flow has the characteristics of suddenness, fast flow speed and strong destructive force. It is of great practical significance and theoretical value to study its dynamic mechanism and accurately simulate the formation process. This model was created in the framework of a Godunov-type finite volume scheme to solve the shallow water equations (SWEs). At the same time, the introduction of Graphics Processing Unit (GPU) improved computing technology greatly, and increased the computing speed. Taking the dam-breaking flume debris flow movement characteristics and the prediction of the risk range of the landslide and debris flow in Pusa Village, Guizhou Province as a simulation reference, the constructed GAST model was used to complete the simulation of the debris flow tank test and the prediction of the risk range. The results show that the numerical simulation is in good agreement with the measured results; the application of GPU acceleration technology effectively solves the large amount of calculation in the process of debris flow simulation problem; and the model runs fast and efficiently on the PC. The model is an ideal tool to study the sudden debris flow process, and can provide timely and accurate decision support for disaster prediction.