The reconfiguration of the entraining-detraining plume model for application in large-scale cumulus parameterization is presented. This procedure extends the entraining-plume model employed in the modified Arakawa-Schubert scheme while considering the detrainment at many levels of the cloud on the basis of the concept of the buoyancy-sorting mechanism of Raymond and Blyth. According to the buoyancy-sorting mechanism, the detrainment results from the neutral buoyancy of the mixture produced through mixing of the cloud air with the ambient air. It has been thought that more entrainment tends to produce more neutral buoyant mixtures because the entrainment of environment air, which is drier than the cloud air usually, causes reducing or losing their buoyancy. Thus, we simply assume that the rate of detrainment to be proportional to that of the entrainment. Of course, more entrainment at a level may not mean more mixtures with neutral buoyancy at that level because the mixture can have neutral buoyancy through the vertical displacement. However, our assumption is reasonable for the large-scale model in considering its large interval between the vertical levels. The arbitrary constant ratio between entrainment and detrainment is determined through a series of sensitivity tests rather than on a physical basis. The properties of the cloud air detrained into the environment are specified from both bulk averaged properties of the updraft cloud air rising from the below and the entrained ambient air. For the validation of the results based on the parameterization, both the single-column model and three-dimensional large-scale model are used. For the simulation of the single-column model, the Southern Great Plains data of the Atmospheric Radiation Measurement project and the Global Atmospheric Research Programs Atlantic Tropical Experiment Phase III data are used.