Cooling drag has traditionally proven to be a difficult flow phenomenon to predict using computational fluid dynamics. With the advent of grille shutter systems, the need to accurately pre-dict this quantity during vehicle development has become more pressing. A comprehensive study is presented in the paper of three automotive models with different cool-ing drag deltas using the commercial CFD solver STARCCM+. The notchback DrivAer model with under-hood cooling provides a popular academic benchmark alongside two fully-engineered production cars; a large saloon (Jaguar XJ) and an SUV (Land Rover Range Rover). Previous studies detail the differences in the flow field; highlighting the interaction between the exiting under-hood cooling flow, and the wheel and base wakes for open and closed grilles. In this study three levels of spatial discretization were used for each vehicle in order to study the importance of accurately capturing the base wake on the absolute and cooling delta drag values and the cooling air mass flow rates. This study is performed using three steady-state RANS solvers (k-ɛ realizable, k-ω SST and Spalart-Allmaras), and the unsteady k-ω SST Detached-Eddy-Simulation. Results show that it is very important to capture both separation and large wake structures in order to recover physically realistic behavior. The RANS models perform well (within 0.005 Cd, 5 counts) on saloon based models, with the k-ɛ realizable model displaying mesh independence. For the SUV model the RANS models predict the correct cooling deltas; however, only the k-ω SST model gives accurate absolute values, with those for k-e realizable and Spalart-Allmaras 22 and 18 counts too high, respectively. The k-ω SST model on the finest mesh contains oscillations in the flow field, particularly in the wake, which are attributable to the unsteady nature of the flow. When averaging the steady-state simulations over 1000 iterations the resulting wake structure is shown to be in close agreement to the unsteady Detached-Eddy-Simulations. The DES model confirms that the variance in the residuals for the k-w SST was indicative of flow unsteadiness.