Structural and thermodynamic properties of elemental and bimetallic nanoclusters are studied at the atomic scale. The modelling is achieved by means of molecular dynamics (MD) and Metropolis Monte Carlo (MC) sampling in the so-called transmutational ensemble. The cohesion model used is based on the second moment approximation of the tight binding model. Copper elemental and Ni x Al 1−x binary alloy clusters are selected as case studies. Particles containing less than n = 201 atoms are predicted to be structureless, except when elemental, formed by n = 13, 55, 135 and 147 atoms. These so-called magic numbers allow icosahedral geometry. Binding energies are not found to be significantly dependent on morphology, suggesting the coexistence of several isomers. As far as Ni x Al 1−x clusters are concerned, phase stability is systematically studied as a function of x, ranging from 0 to 1 and discussed with reference to the bulk ordered alloy. Except in one special case, and in contrast to elemental clusters, no stable phase at all is found in the smallest clusters (n < 201) as they are structureless. In the larger ones, consistently with a recent study with another cohesion model (Campillo J M, Ramos de Dibiaggi S and Caro A 1999 J. Mater. Res. 14 2849), a partition shows up between a core where the bulk stable L1 2 and B2 phases are retrieved and a mantle which may be subjected to aluminium segregation. In the range of cluster sizes considered (n = 13-10 000), the results suggest that, because of the easy surface segregation, the martensitic metastable phase occurring in bulk Ni-Al systems does not take place in free clusters. The segregation efficiency is found to decrease with increasing cluster size while the relative mantle thickness is size independent. This may be the reason why the martensitic phase only occurs in systems larger than currently investigated.