Composite materials fabricated from non-crimp fabric (NCF) are showing potential in primary aerospace components and structures. They can be used in common RFI, RIFT, and RTM manufacturing processes, and are showing promise to replace traditional pre-preg systems. The composite designer hence needs a technique, which can be used to predict the elastic properties of the NCF laminate including manufacturing defects from these infusion processes. The present study develops a representative unit cell (RUC) model of the NCF blanket, which can include potential variability in the NCF structure. Representative unit cells have been developed to predict the two- and three-dimensional mechanical properties in the bi-axial NCF composite. The model is based on the two-dimensional methods of Naik and Scida for woven fiber composites, incorporating the elastic anisotropy theory of Lekhnitskii to extend the predictions to three dimensions. Continuous tow layers with uniform thickness and fiber volume fraction are assumed, and the effects of crimp due to stitching and surrounding resin regions are simulated. The input data are based on measurements and observations of a ±45° carbon-epoxy NCF composite having a 63% fiber volume fraction and less than 1% voids. Output from the model is obtained in terms of the stiffness and compliance matrices as well as traditional engineering properties. For the tensile and compressive moduli, comparisons with data from the carbon/epoxy NCF composite show the values are over-predicted by 11.5—11.6% in tension and 13.5—13.9% in compression. Improved predictions are obtained using a simple stiffness model that allows for different behaviors of the resin as continuous layers or gaps between tows. The resin layers, compared to the resin as gaps, significantly decrease the in-plane moduli, and the model is considered to simulate local through-thickness shear deformation in the resin layers.