In order to quantitatively assess the landslide hazard of a large mountainous area, this study proposes a slope-unit-based probabilistic methodology that involves a three-dimensional (3D) deterministic model coupled with a Geographical Information System (GIS). A study area is divided firstly into slope units which are used as study objects of the 3D deterministic model. For each slope unit, a GIS-based 3D slope stability analysis model is applied to calculate the distribution of 3D safety factors through a Monte Carlo simulation of the slip surface. Assuming the resultant 3D safety factors follow a log-normal distribution, a widely-used reliability index, β, is used to quantitatively evaluate the probability of unsatisfactory performance of each slope unit. An advanced GIS system is developed to efficiently implement all of the computational procedures. The effectiveness of the proposed method has been tested in a practical application to a landslide-prone area in Japan.