Compared to Reynolds averaged NavierStokes (RANS)-based combustion modeling, the large eddy simulation (LES) technique has recently emerged as a more accurate and very adaptable technique in terms of handling complex turbulent interactions in combustion modeling problems. In this article, application of the LES-based combustion modeling technique and the validation of models in non-premixed and premixed situations are considered. Two well-defined experimental configurations where high-quality data are available for validation are considered as case studies to demonstrate the methods, accuracy, and capability of the LES combustion modeling technique as a predictive tool. The large eddy simulation technique for modeling flow and turbulence is based on the solution of governing equations for continuity and momentum in a structured Cartesian grid arrangement. A Smagorinsky eddy viscosity model with a localized dynamic procedure is used as the subgrid-scale turbulence model. A swirl flame is considered as the non-premixed combustion application. For non-premixed combustion modeling a conserved scalar mixture fraction-based steady laminar flamelet model is used. A radiation model incorporating the discrete transfer method is also included in the non-premixed swirl flame calculations. For premixed combustion where the application considered here is flame propagation in a confined explosion chamber, a model based on dynamic flame surface density is used. It is shown that in both cases LES-based combustion models perform remarkably well and results agree well with the experimental data.