In this study the mathematical modelling of a catalytic microstructured plate reactor for the production of hydrogen was performed in 2D and 3D geometry. The proposed reacting system uses the heat generated by an exothermic reaction (combustion) to sustain endothermic reforming reactions. Therefore, it pertains to those devices useful for producing the feed for fuel cell system for the remote generation of electrical power. However, because of the compactness of the reacting system it can also be considered in the context of apparatus aiming at process intensification. Within this frame the catalytic contribution of both exothermic and endothermic reactions was modelled considering the classic Langmuir-Hinshelwood surface kinetic theory. The advantage of using a real 3D geometry configuration consists in the possibility of considering the importance of the entering and boundary effects with particular attention to fluid stagnation and heat hot spots. The trade off of such a choice is certainly the huge increase of computing time and/or of the power of the computing facility. With respect to other works performed with similar reactor geometry and reacting systems this does not use simplifying assumptions such as catalyst layers modeled by one-dimensional approach, fully developed laminar flow or transverse heat and mass transfer taken into account through lumped heat and mass transfer coefficients. Results of simulations presented here concentrates on the comparisons between results of: countercurrent (CTC) and concurrent (CNC) flow patterns of the reactant streams; of simulations carried out with 2D and 3D models and of the influence of the thickness of the catalytic layers on the reactor performance. Simulations indicates that CNC flow pattern of reactants streams allows a better performance of the reactor since positive temperature differences between the catalyst layers and the gas in the channels maintain along the whole reactor and, consequently, there are not heat flux inversions, which occur under CTC flow pattern. Results also showed that as concerns an adiabatic reactor, whatever the operating conditions, 2D and 3D models yield substantially the same results. Finally, modelling demonstrated that for a realistic catalyst layer configuration thicknesses larger than 50 µm are useless for enhancing the reactor performance. The feasibility of the model proposed may show its potential in fast and easy implementation of several combustion and reforming fuels so to significantly enhance the performance prediction of real processes.