This paper presents the characteristics and performances of the HCMFD (Hybrid Coarse-Mesh Finite Difference) algorithm for 3-D pin-by-pin reactor core analyses. The HCMFD algorithm has been suggested for an extremely efficient parallel computing of a pin-by-pin neutron diffusion analysis with a pin-level nodal calculation based on a non-linear local-global iterative framework. The fundamental ideas and concepts for the one-node CMFD scheme, in combination with the conventional two-node CMFD scheme using the nodal expansion method, has been described in detail. When the HCMFD algorithm is applied to a 3-D problem, it provides flexibility on the axial mesh refinement so that one can easily optimize the axial mesh size to treat a significant heterogeneity in axial direction. In this paper, some possible variations in axial mesh set up and several numerical strategies for better convergence are discussed. The overall features of the 3-D HCMFD algorithm were identified through analyses of the 3-D EPRI-9R benchmark, and the potential performance of the HCMFD algorithm for a practical pin-by-pin reactor core analysis has been estimated by solving a big PWR problem. Including the sensitivity study regarding the optimization in axial direction, the parallel performances of the 3D HCMFD algorithm were measured and compared with those of a conventional two-node CMFD scheme. It was shown that the HCMFD algorithm has clear advantages over the conventional CMFD approach in 3-D pin-by-pin core analyses. (C) 2018 Elsevier Ltd. All rights reserved.