The performance of two two Gibbs free-energy-based equation-of-state (EOS) mixing rules is tested for the correlation and prediction of liquid-liquid equilibria (LLE). The LLE of ten binary mixtures were examined using the Peng-Robinson EOS combined with the Wong-Sandler and MHV1 mixing rules with two excess free-energy models (NRTL and UNIQUAC). Good agreement with low-pressure experimental data was obtained for nonassociating and self-associating mixtures using temperature-independent parameters. However, in the case of aqueous mixtures that exhibit both a lower and an upper critical solution temperature, it was necessary to use temperature-dependent parameters, as is also true when activity coefficient models are used directly to describe such systems. The Wong-Sandler mixing rule was capable of predicting the high-pressure phase behavior of the systems studied using parameters from correlation of only low-pressure LLE data, while the MHV1 model required the introduction of a binary interaction parameter. Consequently, the MHV1 model was useful for the correlation but not for the prediction of high-pressure LLE from low-pressure data.