Classical molecular dynamics simulations are used to compute the solvation free energy of two pharmaceutical solids, namely ibuprofen and acetaminophen in carbon dioxide (CO2), over the density range of interest in supercritical processes. In order to examine the influence of the solvent model on the resulting free energies, three popular CO2 models (Zhang, EPM2, and TraPPE) are studied. Relatively large discrepancies for the solvation free energy exist between these CO2 models, suggesting that the former is sensitive to the different balances between dispersive and electrostatic forces used in these models. In particular, for the solvation of the highly polar (dipole moment of similar to 5.2 Debye) acetaminophen molecule, such discrepancies are more pronounced than for the moderately polar ibuprofen (dipole moment of similar to 1.6 D) molecule. Since there is an exponential relationship between the solvation free energy and solubility, the choice of the solvent model substantially affects the predicted solubility. For the solubility of the studied solutes, the value obtained using the TraPPE model is the highest, that of the EPM2 model is intermediate, and that of the Zhang model is the lowest. Generally, the simulations results show that the model with the largest quadrupole moment leads to a more negative solvation free energy and a higher solubility over the entire density range. Further, the decomposition of the solvation free energy into contributions stemming from electrostatics and dispersion interactions shows that the electrostatic interactions are important for a quantitative prediction of solid solubility, while the Lennard-Jones parameters of the solute and solvent are more important for qualitative agreement. Additionally, the infinite-dilution partial molar volume of the two solutes is estimated from the pressure derivative of the solvation free energies. With density increasing beyond the value corresponding to the zero partial molar volume of the solute (minimum solvation free energy), the repulsive part of Lennard-Jones potential wins over the attractive interactions, and the solvent strength of supercritical CO2 decreases; however, due to the increase in the chemical potential of the pure solid (effect of the Poynting correction), the solubility further increases. Overall, these results demonstrate the importance of a proper choice of quadrupole moment of the solvent model, which is crucial for quantitative predictions of the solid solubility in supercritical CO2. (C) 2015 Published by Elsevier B.V.