A new approach based on the statistical associating fluid theory is presented here to model eight light crude oils, with the saturate, aromatic, resin, and asphaltene analysis as the only input for the model. Within the characterization procedure of Punnapala and Vargas (Fuel 2013, 108, 417-429, 10.1016/j.fue1.2012.12.058), the aromaticity parameter and the asphaltene molecular weight were fixed to all crude oil samples, while the asphaltene aromaticity is the only fitted parameter of the model. A correlation for this parameter with the flashed gas molecular weight allows for full predictions of the phase behavior without the need of any asphaltene onset data. The predictive molecular model was used to study asphaltene instability as a function of injected CO2 and natural gas concentration. The model can also accurately reproduce routine pressure volume temperature experiments, such as constant composition expansion, differential vaporization, and multi-stage separation tests, performed on the crude oils, thereby providing a unified framework for phase behavior studies.