A segment-based Eyring-NRTL viscosity model was recently presented and evaluated using data from a variety of mixtures containing polymers. In this paper, the segment-based Eyring-NRTL viscosity model is modified to provide a practical theory-based model for the viscosity of liquid mixtures containing complex components that cannot be typically described by the segment-based approach. To reflect these changes, this new model is referred to as an entity-based Eyring-NRTL viscosity model, because entities are viewed as abstract segments that represent complex components but are not well-defined. The entity-based Eyring-NRTL viscosity model provides a theory-based equation for liquid mixture viscosity as a function of composition, temperature, pressure, and shear rate. The correlative and predictive capability of the entity-based Eyring- NRTL viscosity model is demonstrated using component and mixture viscosity data from the following systems: lubricating oil blends, crude oil blends, bitumen-diluent mixtures, heavy oil-diluent mixtures, and acetone-ethanol-2,2,4-trimethylpentane mixtures. Ternary lubricating oil blend viscosities and ternary acetone-ethanol-2,2,4-trimethylpentane mixture viscosities are predicted from component and binary viscosity parameters determined from only component and binary mixture viscosity data. The entity-based model should be useful for modeling viscosity in a number of applications containing complex components, such as petroleum, petrochemicals, polymers, lubricants, coatings, and personal care products.