Computational prediction of separated flows is an area of interest, specifically in applications to gas turbine engines, liquid pumps, and many other engineering applications. Although these types of flows are governed by the Navier-Stokes equations, direct numerical simulation (DNS) of practical engineering flows is currently too expensive in terms of the required computational time. It is therefore a case to attempt simulation of these flows using the Reynolds-Averaged Navier-Stokes (RANS) equations which must be closed by utilizing a turbulence closure model.;The present study used the NASA Glenn-HT code, a compressible Navier-Stokes solver, with two different turbulence models, the u2 --f model of Durbin and the zeta-- f model of Hanjalic, to observe their abilities to predict separated flows. The elliptic relaxation turbulence models and their implementation in Glenn-HT are described. Three cases are described to show the ability and limitations of these turbulence models.