Association theory has been extensively applied to equations of state such as statistical associating fluid theory and cubic plus association. This work presents an integration of association theory with the nonrandom two-liquid segment activity coefficient (NRTL-SAC) model. Replacing the four conceptual segments (hydrophobic, hydrophilic, polar attractive, and polar repulsive) in the original NRTL-SAC model, the association NRTL-SAC model represents interacting molecules with two conceptual segments (nonpolar and polar) and two association site types (hydrogen bond acceptor and donor). Given the molecule-specific parameters, the model shows much improved predictive capability than the original NRTL-SAC model for phase equilibria of chemical mixtures in which associations take place.