This study simulates soot formation processes in diesel combustion using a large eddy simulation (LES) model, based on a one-equation subgrid turbulent kinetic energy model. This approach was implemented in the KIVA4 code, and used to model diesel spray combustion within a constant volume chamber. The combustion model uses a direct integration approach with a fast explicit ordinary differential equation (ODE) solver, and is additionally parallelized using OpenMP. The soot mass production within each computation cell was determined using a phenomenological soot formation model developed by Waseda University. This model was combined with the LES code mentioned above, and included the following important steps: particle inception during which acenaphthylene (A2R5) grows irreversibly to form soot; surface growth with driven by reactions with C2H2; surface oxidation by OH radical and O2 attack; and particle coagulation.The results obtained using our new model are compared to those generated using a RANS (RNG k-epsilon) model, and also to experimental data from the engine combustion network (ECN) of Sandia National Laboratories. The sensitivity of the LES results to mesh resolution is also discussed. The results show that both RANS and LES simulations predict the dispersion and vapor penetration of the injected fuel fairly well. LES generally provides flow and spray characteristics in better agreement with experimental data than RANS. It is also shown that the phenomenological soot model is useful for investigating soot particle production and distribution. The LES model was better than the RANS model at describing instantaneous soot concentration contour.