Extremely-rapid flowlike landslide are a major hazard in many parts of the world, and managing their risk requires an understanding of the mechanisms that drive motion, as well as reliable predictions of their potential destructiveness. Numerical runout models are one tool that can be used for both these applications, however their utility is presently limited by their computational runtime. In the present work, a new depth-averaged, smooth particle hydrodynamics based model is implemented to run on a graphical processing unit. The new implementation provides a speedup of over two orders of magnitude, compared to a commonly used CPU based runout model. The new model has been validated, and then is used to back-analyse the Johnsons Landing Landslide. It is shown that increasing model resolution results in an accurate simulation of complex topographic interactions between the flowing landslide and the surface topography. The new model runs on commercially available GPU’s, and should therefore be useful for both researchers and practitioners seeking to understand and quantify landslide hazard and risk.