This study presents an approach for predicting surface tensions (ST) of mixtures coupling the critical-point-based perturbed-chain statistical association fluid theory (CP-PC-SAFT) and the square gradient theory (SGT). This modeling framework implements standardized and transparent parametrization procedures. As an input it requires the pure compound critical temperatures and pressures along with the triple point densities and surface tensions, without further adjustment to ST data of mixtures. Its predictions of a major part of the existing surface tension data on asymmetric binary systems, including the aqueous solutions, in the entire available range of conditions, are discussed. The results demonstrate that the proposed approach is characterized by a remarkable overall robustness and reliability. However, the system toluene (l)-water (2) poses a clear limitation to its predictive implementation. Predictions of CP-PC-SAFT coupled with the generalized modified YarrantonSatyro correlation for the recently published data on saturated liquid CO2 (1)-n-alkane (2) phases are also presented.