Clusters of TeF6 ranging in size from 100 to 350 molecules spontaneously transformed from body-centered to monoclinic structures as they were cooled. To monitor the extent of the nonreconstructive phase change, it was necessary to identify molecular reorientations rather than to analyze Voronoi polyhedra. The strong dependence of configurational energies per molecule upon cluster size could be reproduced by a simple model attributing the entire effect to the surface layer of molecules. Without evidence that core molecules in clusters behave like those in the bulk, there would have been no basis for estimating the degree of supercooling. Nucleation of the monoclinic phase always took place in the cluster cores. Single crystals invariably resulted from the transformation, in accord with Kashchiev’s criterion for mononuclear transitions based on growth rates of the new phase. Simulations yielded a crude estimate of the nucleation rate from which the interfacial free energy of the boundary between the monoclinic and bcc phases could be derived via classical nucleation theory. The ratio between the interfacial free energy and heat of transition was found to be similar to that reported by Turnbull for liquid-solid interfaces. Two alternative procedures to determine the sizes of critical nuclei in the clusters were in rough agreement with each other but of marginal physical significance because, to date, insufficient attention has been paid to the apportionment of molecules in the transition layer between the old and new phases. Nor has the effect of the transition layer between the phases been adequately taken into account in the classical theory applied herein. Nevertheless, it is clear for the highly supercooled clusters of the present investigation that the structural fluctuations leading to the critical nuclei are only a few molecules across.