A numerical model for turbulent reacting flow is described and applied for predictions in an industrial gas turbine combustor operating on low-Btu coal gas. The model, based on fast-reaction limit, used Favre averaged conservation equations with the standard k-c model of turbulence. Effects of turbulent fluctuations on chemistry are described statistically in terms of the mean, variance and probability density function (assumed to be fldistribution) of the mixture fraction.Two types of geometric approximations, namely axisymmetric and three-dimensional, were used to model the combustor. Computations were performed with (a) no swirl (b) weak swirl and (c) strong swirl at the fuel and primary air inlets. Essentially, the same bulk mean temperature distributions were obtained for axisymmetric and three-dimensional calculations while the computed pattern factors and the liner wall temperatures for the two differed significantly. Complete combustion was predicted with strong swirl, a result supported by the available test data. The maximum liner wall temperature predicted for three-dimensional calculations compared favorably with the experimental data while the predicted maximum exhaust gas temperature differed by =120 K . The difference was attributed to measurement uncertainties, model assumptions and lack of accurate data at the inlets. The maximum flame temperature was below 1,850 K indicating that thermal NOx may be insignificant.