A three-dimensional multi-zone combustion model is developed for the description of the combustion mechanism inside the engine cylinder of direct injection diesel engines. Various multi-zone models have been proposed in the past for the prediction of DI diesel engine performance and emissions. These models offer an alternative tool if one wants to avoid the use of other more complicated and sophisticated flow models that require high computational times. Most of them have the disadvantage that they focus mainly on emissions, failing to predict at the same time engine performance adequately. In almost all multi-zone models the resulting fuel jet after injection, which is divided into zones, is assumed to be symmetrical around its axis. In the present work a different approach is followed. The fuel jet is divided into zones in the three dimensions overcoming the need for the previous symmetry assumption. This results to an increase of the number of zones involved and thus computational time, but it offers an insight to the combustion and pollutants formation mechanism in great detail. Using the 3-D multi-zone model we are in a position to describe the local pollutants formation history inside the fuel jet and to detect the mechanisms that lead to their formation. From the computational results it is revealed that the symmetry assumption around the jet axis is not valid, especially in the presence of air swirl where the distribution of thermodynamic properties, air and fuel concentration around the jet axis is highly non-symmetric. To validate the model an experimental investigation has been conducted on a Lister LV1 series, direct injection, single cylinder, test engine located at the authors laboratory. The comparison of experimental findings and computational results reveals a very good degree of agreement for both engine performance and pollutants emissions. This is encouraging, especially if we take into account the complexity of the proposed model. Specifically, the computational findings for pollutants compare better with experimental values when compared with a previous two-dimensional multi-zone model developed by the present research group.