Several predictive GE methods based on the group-contribution concept have been tested for prediction of multicomponent vapor–liquid equilibria (VLE) of fragrance mixtures. The evaluation covered 24 binary, 20 ternary, and 21 quaternary mixtures combining six different fragrances typically used in perfume formulation and ethanol as the solvent. The selected predictive methods are the original UNIFAC (updated to the last revision), the UNIFAC-Dortmund, the ASOG, and the A-UNIFAC. The equilibrium compositions were experimentally measured at 23 °C by headspace gas chromatography (HS-GC). For each liquid composition, the corresponding experimental and predicted vapor compositions were compared. This study confirmed that the UNIFAC method is the best predictive model for prediction of VLE, while the A-UNIFAC performed poorer than other evaluated methods for the mixtures studied in this work. Moreover, from the vapor composition the odor intensity and character of the fragrance mixtures was calculated using a previous developed model which considers the Stevens’ power law for the olfaction intensity scale and the stronger component model for quality perception. In this way, the odor intensity and character of fragrance mixtures was predicted and compared to experimental and sensorial evaluations. From this work it can be concluded that the UNIFAC and UNIFAC-Dortmund methods allow obtaining optimal predictions of the character of the fragrance mixtures with an agreement of 95.4% when compared to experimental measurements.