A multi-zone model for the prediction of performance and emissions of indirect injection diesel engines is presented. The interaction between turbulent mixing and chemical reaction, which primarily controls combustion, is described as a stochastic process by means of a Monte Carlo collision-dispersion model. The major contribution of the present work lies in combining such an approach with improved models of fuel evaporation in supercritical conditions, burning rate, radiative heat transfer and flow coefficient of the passageway, besides properly accounting for the real gas effects. A limited number of empirical correlations was introduced, so obtaining a rather general and physically meaningful predictive tool.Experimental data concerning a small single-cylinder engine have been employed to validate the computational procedure. After calibrating the model in a single operating condition of the engine, the predicted and measured performance and emissions have been successfully compared at several engine speeds and loads. The computational cost of the simulation appears to be acceptable, although the results have to be averaged over at least five engine cycles to smooth away the random fluctuations due to the stochastic character of the model.