The feasibility of a trained version of the Jouyban-Acree ( J-A) model was evaluated in this research to predict the solubility of ketoprofen (KTP) in some propylene glycol + water cosolvent mixtures. The usefulness of the solubility log-linear equation proposed by Yalkowsky and Roseman (Y-R) was also evaluated for the same drug in this cosolvent system. The solubility estimation was studied as a function of temperature and cosolvent composition. The J-A and Y-R models require only the experimental equilibrium solubility values in the pure solvents at all the temperatures evaluated. The estimated solubility values obtained by using both semiempiric models were similar between them (differing in 0.05 natural logarithmic units), but they present notorious deviations with respect to the experimental values, in particular the Y-R model. These results demonstrated that it is necessary to improve the theoretical strategies for estimating this property, and on the other hand, they also demonstrated the great importance of the experimental determination of drugs solubility in those cosolvent mixtures useful in dosage forms design.