A new electrolyte model, which is based on the predictive thermodynamic model COSMO-RS, is presented. For this purpose, an implementation of COSMO-RS that allows the integration of multiple segment descriptors was developed. To aid in the development of the electrolyte model, a new technique is presented that allows the evaluation of the different contributions of the interaction terms of COSMO-RS to the partial molar enthalpies. General empirical interaction energy equations are introduced into the electrolyte model. They are parameterized based on a large training set of mean ionic activity coefficients as well as liquid-liquid equilibrium data close to ambient conditions. The model is shown to be capable of predicting properties of systems containing anions that were not part of the training set of the model. Furthermore, it is demonstrated that the model can also lead to satisfying predictions if compared to vapor-liquid equilibrium data. (c) 2017 American Institute of Chemical Engineers AIChE J, 63: 272-285, 2018