NiTi alloys exhibit diverse mechanical as well as functional properties and are the most commonly used SMAs (shape memory alloys) for engineering as well as medical applications. NiTi’s shape memory effect is caused by a fully reversible martensitic transformation. The transformation is accompanied by macroscopic changes in the material’s stiffness and strain evolution which makes NiTi particularly relevant for sensors, actuators and damping elements. The following work deals with the modeling of the cubic to monoclinic transformation in nano- and polycrystalline NiTi, which is triggered by temperature. It proceeds in a broad temperature interval starting after the very small temperature interval of intermediate transformation to an orthorombic phase at around 30°C. The high-temperature, high-symmetry phase called austenite is a cubic, ordered crystal and the low-temperature, lower symmetry phase called martensite is monoclinic. In order to accommodate the new phase a twinned crystal structure is formed. Using the nonlinear theory of martensitic transformations, starting only from lattice parameters of the cubic and monoclinic phase, the deformation gradients describing the shape changes of all possible martensitic variants and variant-pairs forming a twin are calculated. A nano-grain distribution below 100nm average diameter is modeled, since for these small grain sizes the preferred martensite morphology turns out to consist of a single laminate of alternating variants. The material itself is modeled as a thermoelastic solid. Anisotropic material behavior with elastic constants from ab initio calculations are used in combination with locally random orientations at the grain level. It was found that the elastic strain energy constitutes the main contribution to the total energy barrier. In this work an incremental algorithm for the transformation was developed based on an energy minimizing principle. The so obtained transformation kinetics agrees with the experimental evidence reported in the literature.