Cluster dynamics is a mesoscopic modeling technique describing the various kinetic stages of homogeneous precipitation by the same set of rate equations. However, when the simulated cluster size continuously increases, it easily causes an enormous computational workload, and the use of a particle-size-grouping method is often necessary to solve this problem. In this study, an ungrouped cluster dynamics model and certain existing grouping methods are reviewed. Next, a new grouping method with an assumed logarithmically-linear distribution of cluster number densities inside each group size is proposed. Comparing the results of all grouped models with the exact solution of the ungrouped model for simulating aluminum-scandium (Al₃Sc) precipitation in the Al-0.18%Sc (atomic fraction) alloy at 300^oC, the new grouping method was able to reduce computational costs considerably keeping enough total and local accuracies. Moreover, the reasonable agreements of the mean radii and size distributions as functions of time between experiments and simulations were obtained, demonstrating the ability of the new grouping method in modeling large-scale precipitation kinetics.