In recent work [Peng et al. Fluid Phase Equilibr. 227(2)(2009).131] a molecular-based group contribution SAFT-VR equation of state (GC-SAFT-VR) was proposed as a predictive tool for the study of fluid phase behavior. In the GC-SAFT-VR approach a fluid is modeled by a chain of tangentially bonded segments that each represent the different functional groups in the molecule, such as CH3, CH2 and OH, and so represents a more physically realistic model than that used in other SAFT-based approaches in which the segments of the model chain are identical i.e., have the same model parameters. The molecular parameters of the functional groups are determined by fitting to experimental vapor pressure and saturated liquid density data for selected members of different chemical families in a successive fashion, i.e., alkanes for CH3 and CH2 groups and subsequently alcohols for the OH group. The transferability of the molecular parameters can then be tested by comparing the theoretical predictions with experimental data for the phase behavior of pure fluids not included in the fitting process as well as binary mixtures of both simple fluids and polymer systems. Here, we continue the development of the GC-SAFT-VR approach, and for the first time, test its ability to predict the phase behavior of ternary mixtures. In particular, we consider binary and ternary mixtures of 1-pentanol, dibutyl ether and n-nonane. We find that the GC-SAFT-VR approach provides a good description of the phase equilibria of these binary and ternary systems without the need to fit cross interaction parameters to experimental mixture data. (C) 2010 Published by Elsevier B.V.