Local composition models of newly proposed Wilson solvation (WS) model and preferential solvation (PS) model were developed for predicting Kamlet-Taft (KT) dipolarity/polarizability (π*) of binary and ternary solvent mixtures. The WS model could correlate the π* data of eleven nonpolar (1) - polar (2) liquid mixtures that gave an average relative deviation (ARD) of 5.4%, compared with the PS model (3.7%). The WS and PS parameters obtained from correlations were generalized to have a linear relationship with dipole moment (μ). The resulting linear relationship allows the predicted π* values of 24 additional binary mixtures to within an overall ARD of 9.3%, by knowing only pure component π* and μ values. The WS and PS models were applied to predictions of nonaqueous ternary systems for KT π* values, KT basicity, KT acidity and Reichardt normalized polarity that gave an overall ARD of 5.8% for WS model and 3.3% for PS model. Application of the models makes it is possible to identify favorable compositions of ternary (heptane-ethyl acetate-ethanol) solvents being used in replacements of dichloromethane for chromatography and other chemical ternary systems that require only binary polarity data.