This paper describes the inclusion of variable gas properties within a Reynolds average Navier-Stokes solver for turbomachinery and its application to multistage turbines. Most current turbomachinery computational fluid dynamics (CFD) models the gas as perfect with constant specific heats. However, the specific heat at constant pressure can vary significantly. This is most marked in the turbine where large variations of temperature are combined with variations in the fuel air ratio. In the current model is computed as a function of the local temperature and fuel air ratio using polynomial curve fits to represent the real gas behavior. The importance of variable gas properties is assessed by analyzing a multistage turbine typical of the core stages of a modern aeroengine. This calculation includes large temperature variations due to radial profiles at inlet, the addition of cooling air, and work extraction through the machine. The calculation also includes local variations in fuel air ratio resulting from the inlet profile and the dilution of the mixture by the addition of coolant air. A range of gas models is evaluated. The addition of variable gas properties is shown to have no significant effect on the convergence of the algorithm, and the extra computational costs are modest. The models are compared with emphasis on the parameters of importance in turbine design, such as capacity, work, and efficiency. Overall the effect on turbine performance prediction of including variable gas properties in three-dimensional CFD is found to be small.