Computational Fluid Dynamics (CFD) simulations are widely used in many wind-related studies, including cross-ventilation, in urban areas. The accuracy of the CFD models, however, is still a challenging issue for accurate prediction of the complex flow behavior around and inside the buildings. Application of sophisticated CFD models, such as Large Eddy Simulation (LES) and unsteady Reynolds averaged Navier-Stokes (RANS), are generally limited, so many researchers and designers utilize the steady RANS models for design and analysis of crossventilation performance in urban areas. The RANS models, however, provide poor results in predicting the cross-ventilation in street canyons.Thus, this study aims to understand and quantify limitations of the steady RANS models for cross-ventilation applications in highly-packed urban areas. To this end, a series of CFD simulations were conducted for a group of buildings, which were arranged in regular and staggered orders with different urban area densities. Both sealed-body and cross-ventilated scenarios were considered in this study while the surface-averaged and local values of the wind pressure were compared with the results from a wind tunnel measurement by Tamura (2012).Furthermore, the possibility of the RANS model improvement was considered using a parameter sensitivity study over the closure coefficients of a RANS model. Therefore, new coefficients for urban area with densities between 0.2 and 0.4 were found to significantly improve the accuracy of the RANS model. Nonetheless, as an interesting finding of this study, for higher values of urban area densities above 0.4, CFD results went outside the expected measurement ranges; this implies that CFD modeling of higher density urban areas should be treated with more cautious and further studies are required to develop a guideline for such applications.