The combined computational fluid dynamics and discrete element method (CFD-DEM) approach has been proved to be an effective tool to study the fundamentals of different particle-fluid flow systems, but suffers high computational cost problem. Recently, various treatments such as parcel-particle concept, coarse-grained model, similar particle assembly and representative particle model have been developed to reduce the computational cost of CFD-DEM approaches. These treatments are basically empirical and thus their applicability is likely system-dependent. Until now, there are still no general agreements on the formulation of those models and their accuracy and general applicability are largely unknown. In this work, a coarse-grained (CG) (CFD-DEM) model is developed to model the swirling multiphase flow in a dense medium cyclone (DMC) and the error caused by the CG concept is quantified by carrying out controlled numerical calculations to directly compare the simulated results between a standard CFD-DEM model and a CG CFD-DEM model. It demonstrates that when the flow is dilute, the results are independent on the size of the grain (also called as parcel or model particle in this work). Nonetheless, when the flow is dense, small discrepancies are observed between the two models. This work suggests that the CG CFD-DEM model is indeed a useful tool to quickly evaluate the flow and performance of large-scale DMCs and the simulation results should be useful at least qualitatively, if not quantitatively. (C) 2016 Elsevier Ltd. All rights reserved.