Computational fluid dynamics (CFD) simulations have the potential to provide predictions of flow around airfoils and wings of arbitrary complexity, without the need to perform real-world testing and experimentation. However, the computational cost of such simulations increases dramatically as the complexity of the modeled geometry increases, thus imposing limitations on the use of CFD for design optimization where many different flight conditions must be considered. Reduced-order models can overcome these limitations by providing rapid calculations of the flow field at a fraction of the computational cost, but the accuracy of such models can be substantially reduced in flows with complex physics, such as flow separation. In this paper, we compare the accuracy of three reduced-order models for calculating the coefficient of pressure, Cp, on both simple (i.e., NACA0012 airfoil) and complex (i.e., NACA64A006 wing) aerodynamic configurations at different angles of attack, including high angles of attack where flow separation occurs. These models are trained using CFD data and the capability of the models to predict Cp for new angles of attack is characterized. We find that reduced-order models based on local interpolation methods are the most accurate, although the accuracy becomes worse overall for a threedimensional wing and worse in particular for high angles of attack where flow separation occurs.