In recent years, porous shape memory alloys have found several industrial applications. Thanks to biocompatibility, corrosion resistance, and superior mechanical properties, porous NiTi has been introduced as a promising candidate for being used as bone scaffolds. Since the mechanical response of a scaffold is of importance in order to prevent stress-shielding phenomena and trigger ossteointegration, predicting the mechanical response of these scaffolds before fabrication is inevitable. In this paper, a new mesoscale model based on Voronoi tessellation of three-dimensional space is presented for the simulation of porous shape memory alloys. To do so, after tessellating the space, some cells are selected randomly to be assigned as pores and a suitable constitutive model of dense SMA is attributed to the other cells. The model is validated against experimental findings reported in the literature demonstrating good agreement. In addition, the effects of number of cells, level of randomness, and the type of boundary conditions on the stress–strain response is assessed. The results show that in order to achieve desirable results, the number of cells and the value of randomness must be chosen greater than minimum corresponding values. As another result, the geometrically periodic model is more computationally efficient than the mechanically periodic one.