Abstract: In the loess plateau, shallow loess landslides are especially widespread and frequent geological disasters, causing serious casualties as well as huge property damage. Although two-dimensional deterministic models are widely applied to assess the stability of shallow landslides, they could not sufficiently consider the three-dimensional spatial variation of geotechnical property, layering configurations and groundwater. It might not conform with the actual situation of slope stability. Therefore, the three-dimensional deterministic model with considering complicated slope situation has the great significance to acquire the results that are more accordant with the actual situation. At the same time, it will exercise a profound and far-reaching influence to effectively mitigate landslide disasters. This paper takes the three-dimensional deterministic model Scoops3D to evaluate its adaptation and reliability of predicting shallow loess landslide stability. Firstly, the sensitivity analysis of model calculating parameters indicates that the most influential parameters on accuracy of safety factor are cohesive force, sliding direction of visual angle and the weight of grids, so it could guide to acquire the detail key input parameters. Then, the different resolution digital elevation models (DEMs) and geotechnical parameters are selected and used to predict the stability of shallow loess landslides in the typical gully and ridge physiographic region by using the Scoops3D. Comparing the calculated results with the detail inventory of point landslides and facial shape landslides shows that this model has a high accuracy in predicting shallow landslide stability. At the same time, the inventory of point landslides may be more suitable to model verification than facial shape landslides. Finally, the confusion matrix and the success rate curve are used to examine the reliability of predicted results that based on different resolution DEMs. The results prove that this model has a good adaptation to predict the stability of shallow loess landslides in the selected study area. Meanwhile, it could obtain reliable prediction accuracy with the high-resolution DEM data.