A new weighted-sum-of-gray gases (WSGG) model that is based on the statistical narrow band model (SNB) RADCAL is proposed for use in computational fluid dynamic (CFD) simulations of air and oxy-combustion. When employed in conjunction with the discrete ordinates (DO) method, the model predictions compare well against line-by-line benchmark data that have been made available recently that are based on the latest spectroscopic databases. Furthermore, the model compares well against the EM2C SNB model calculations that have served as benchmark data in three-dimensional geometries. Radiative transfer calculations in these prototypical problems therefore confirm recent experimental observations that SNB RADCAL and EM2C SNB serve as good model databases to develop approximate radiative property models. To achieve an optimum balance of speed and accuracy in computationally intensive CFD simulations, non-gray formulations of the WSGG model are also employed with the P1 model and solutions are compared against those generated by the DO model. While the P1 model gave favorable comparisons when cold, black walls were present, the errors in the surface incident radiative flux predictions increased in the presence of hot, reflecting walls. Finally, in fully coupled simulations of natural gas combustion under air-firing and oxy-firing modes, the predicted incident radiative flux profiles were distinctly different between the gray and non-gray calculations at regions of high temperature gradients, while the centerline temperature predictions were comparatively unaffected. The effects of turbulence radiation interactions were also accounted for through the temperature self-correlation term. However, the magnitudes of the temperature fluctuations were small and localized within this furnace and did not significantly alter our predictions. Copyright (c) 2012 John Wiley & Sons, Ltd.