Computational modeling is of great importance in solvent selection for new active pharmaceutical ingredients (APIs), with the Perturbed-Chain Statistical Associating Fluid Theory (PC-SAFT) equation of state being among the most popular tools for modeling the API solubility. The PC-SAFT parameters for APIs are traditionally fitted to experimental solubility data, leaving the PC-SAFT performance for other thermodynamic properties of pure APIs and API-solvent mixtures unknown. Therefore, the intention of this study was to investigate the PC-SAFT performance for the solubility as well as pure component properties (liquid density and vapor pressure) of five model APIs: paracetamol, ibuprofen, naproxen, indomethacin, and dibenzofuran. To this end, five different parametrization strategies were defined, the corresponding new parameter sets were identified (using the simulated annealing technique), and their impact on the PC-SAFT performance was evaluated. These strategies differed mainly in the combination of properties included in the parameter regression. The results showed that the API parameters fitted only to solubility data provided a very poor estimation of the pure API properties, whereas those fitted to the liquid density and vapor pressure provided not only an accurate description of such properties but, in many cases, solubility predictions comparable to those obtained using parameters based merely on the solubility. It was also revealed that the inclusion of the vapor pressure in addition to solubility improved the solubility prediction for API-solvent systems not included in the parameter regression. Moreover, the effect of explicitly accounting for the API dipole moment in the PC-SAFT framework was examined.