Infinite dilution activity coefficients (IDACs) are important characteristics of mixtures because of their ability to predict operating behavior in distillation processes. Thermodynamic models are used to predict IDACs since experimental data can be difficult and costly to obtain. The models most often employed for predictive purposes are the “original and modified UNIFAC Group Contribution Methods” (GCMs). COSMO-RS (COnductor-like Screening MOdel for Real Solvents) is an alternative predictive method for a wide variety of systems that requires a limited minimum number of input parameters. A significant difference between GCMs and COSMO-RS is that a given GCMs' predictive ability is dependent on the availability of group interaction parameters, whereas COSMO-RS is only limited by the availability of individual component parameters. In this study COSMO-RS was used to predict infinite dilution activity coefficients. The database assembled by, and calculations with various UNIFAC models carried out by, Voutsas and Tassios (Ind. Eng. Chem. Res. 1996, 35, 1438) were used as the basis for this comparison. This database contains aqueous and nonaqueous nonelectrolyte binary mixtures. In aqueous systems, COSMO-RS provides the best result for systems with alkyl halides or aromatics as solutes in water; all other binary mixtures also gave good results. Overall, COSMO-RS was less successful for nonaqueous systems. Good results from COSMO-RS were obtained for nonaqueous systems involving alkyl halides, alkanes in ethanol, and ketones in alkanes.