Numerical values of the NRTL equation parameters for calculation of the vapor liquid liquid equilibria (VLLE) at atmospheric pressures for 27 ternary mixtures are presented. These values were fitted to the experimental VLLE and vapor-liquid equilibrium (VLE) data to describe simultaneously, as accurately as possible, VLE and liquid liquid equilibria (LLE). Coefficients obtained in this manner allow calculation of the VLLE and VLE of ternary mixtures and binary subsystems with sufficient accuracy. In the case of VLLE, the model, called NRTL-VLL, usually gives mean deviations between the calculated and measured component concentrations of less than 1.3 mol %. Average deviations of the calculated and measured temperatures (or pressures) for most of the mixtures did not exceed 0.45 K (or 3%). Simultaneously, the model calculates the VLE of the homogeneous part of the considered ternary systems and their binary subsystems with good or fairly good accuracy. VLLE calculations were also carried out for four models based on equations of state (EoS). They predicted the VLLE much less accurately than the NRTL-VLL model. For VLLE calculations at high pressures (six systems), the Wong and Sandler model was used in which the coefficients k(ij) were equal to 0 and the G(E) value was calculated with the NRTL equation. The parameters of this equation were fitted to the experimental VLLE data. This model, called WS-NRTL, describes VLLE with high accuracy. In most cases, the mean deviations between the calculated and measured component concentrations in the gas phase and the two liquid phases did not exceed 1 mol %, and the average deviations between the calculated and measured pressures were lower than 3%. For two mixtures at high pressure, the another method presented in the literature was also tested. This method used the Wong-Sandler model in which G(E) was calculated by the UNIFAC method and the coefficients k(ij) in the mixing rule were fitted to the experimental binary VLE data. This model showed a quite good accuracy but was less precise than the presented WS-NRTL method.