The study was aimed at quantitative comparison of retention properties of modern stationary phases for reversed-phase HPLC. Three approaches, the calculated logarithm of octanol/water partition coefficient (clogP)-based model, the molecular modeling descriptors-based model and the hydrophobic-subtraction model, were compared and discussed. Gradient retention time, tR, of a series of test analytes was a dependent variable in the quantitative structure-retention relationship (QSRR) equations describing retention in terms of analytes' structure descriptors. The QSRRs derived were used to characterize in quantitative manner the specific retention properties of nine representative reversed-phase HPLC. Either the theoretically calculated logarithm of octanol/water partition coefficient, or the structural descriptors from molecular modeling were employed to quantitatively characterize the structure of the analytes. The three molecular modeling-derived structural descriptors considered were: the total dipole moment, the electron excess charge of the most negatively charged atom and the water-accessible molecular surface area. In addition to the above standard QSRR approaches, a recently developed parameterization of reversed-phase column selectivity based on the hydrophobic-subtraction model of Snyder et al. [L.R. Snyder, J.W. Dolan, J.W. Carr, The hydrophobic-subtraction model of reversed-phase column selectivity, J. Chromatogr. A 1060 (2004) 77] was considered. According to the hydrophobic-subtraction model, reversed-phase columns are characterized by five selectivity parameters derived from the linear solvation energy relationships (LSER) theory. Values of these parameters are available for more than 300 different columns. It has been demonstrated that the clogP-based model, the molecular modeling descriptors-based model and the hydrophobic-subtraction model provide generally similar classification of the HPLC columns studied. Some differences in column classification by the three approaches considered are discussed in terms of specific properties of individual stationary phases. All the approaches allow a quantitative, although multidimensional, characteristic of HPLC columns, however, the nonempirical QSRR-based approach is simpler and require less labor.