Combustion models can have a significant impact on flame simulations. While solving finite rate chemistry typically yields more accurate predictions, they depend significantly on the detailed kinetics mechanism used. To demonstrate the effect, Large Eddy Simulation (LES) of Sandia Flame D  has been performed using various combustion models. Four different detailed kinetics mechanisms have been considered. They include DRM mechanism with 22 species, GRI-mech 2.11 with 49 species, GRI-mech 3.0 with 53 species , and Model Fuel Library (MFL) mechanism with 29 species . In addition to the mechanisms, two modeling approaches considered are direct integration of finite rate kinetics (FR) and Flamelet Generated Manifold (FGM). The performance is compared between combinations of the mechanisms and combustion-modeling approaches for prediction of the flame structure and pollutants, including NO and CO. The mesh contains about half a million hexahedral cells and LES statistics were collected over ten flow throughs. Advanced solvers including dynamic cell clustering using the Chemkin-CFD solver in Fluent have been used for faster simulation time. Based on comparison of simulation results to the measurements at various axial and radial positions, we find that the results using the FGM approach were comparable to those using direct integration of FR chemistry, except for NO. In general, the simulation results are in good agreement with the experiment in terms of aerodynamics, mixture fraction and temperature profiles. However, kinetics mechanisms were found to have the most pronounced effect on emissions predictions. NO was especially more sensitive to the kinetics mechanism. Both versions of the GRI-mech fell short in predicting emissions. Overall, the MFL mechanism was found to yield the closest match with the data for flame structure, CO, and NO.