The aim of this work is to evaluate the performances of two excess Gibbs energy models, namely the Scatchard-Hildebrand model and the combined Scatchard-Hildebrand/Flory-Huggins model, in correlating the composition and temperature dependence of the solubility of diazepam in water + tert-butyl alcohol solvent mixtures. The dependence of the pure component properties required as input data to the models on temperature was considered as well as that of the adjustable binary interaction parameters. For this purpose, a set of twenty-seven model versions containing from three up to six adjustable parameters and differing from one another by the dependence of at least one binary interaction parameter on temperature was generated for the two excess Gibbs energy models investigated. To rank and weight among these different model versions, second-order Akaike's information criterion corrected for small sample size was used. The correlative performances of the most parsimonious versions of the Scatchard-Hildebrand and combined Scatchard-Hildebrand/Flory-Huggins models selected from this approach were then evaluated, compared and discussed. (C) 2017 Elsevier B.V. All rights reserved.