This study compares the performance of COSMO-SAC-Phi (CSP) and other predictive models from literature when calculating the solubility for 2450 experimental data points with pressures ranging from 45 to 500 bar and temperatures varying from 298 to 473 K. For the calibration of the pure compound CSP model parameters, pure compound saturation pressure and liquid volume were used, no binary parameters were adjusted. The average logarithm deviation in mole fraction was found to be 0.46 and 0.51 for the CSP (2021) and CSP (2019) versions, respectively. The CSP (2021) results represent an overall improvement of 27 %, 48 %, and 64 % when compared to PSRK, PRCS, and SRK-MC+vdW results, respectively. This indicates that CSP can give useful solubility estimates, especially when no experimental solubility data are available.