An extensive infinite-dilution partial molar volume (PMV) data base of various liquid and solid solutes in supercritical carbon dioxide, including the very recent experimental partial molar volume data reported by Spicka et al. [B. Spicka, A. Cortesi, M. Fermeglia, I. Kikic, Determination of partial molar volumes at infinite dilution using SFC technique, J. Supercrit Fluids 7 (1994) 171] and Cortesi et al. [A. Cortesi, I. Kikic, B. Spicka, K. Magoulas, D. Tassios, Determination of partial molar volumes at infinite dilution of alcohols and terpenes in supercritical carbon dioxide, J. Supercrit. Fluids 9 (1996) 141] is used for a comprehensive study of the capabilities of three different thermodynamic models for the prediction of infinite-dilution partial molar volume data. The models are two equations of state, i.e. the well-known Peng-Robinson cubic EoS and the perturbed hard-chain theory (PHCT) EoS, and an EoS/G(E) model, the recently proposed LCVM group-contribution model. Correlation of the whole data base has been also considered with the Peng-Robinson EoS in order to determine the capability of a cubic equation of state to describe PMV data. Generally, all models were found to be semi-quantitative in the PMV prediction, especially in the range of the PMV minimum. Satisfactory prediction results are obtained using the Peng-Robinson EoS with interaction parameters from phase equilibrium data, and the fully predictive LCVM model. Both LCVM and the PHCT models tend to overpredict the PMV of n-alkanes in carbon dioxide in the high-pressure region, while PHCT always gives less satisfactory prediction results.