The structures of various compositions of nanoalloy clusters CuxAuy (x + y less than or equal to 30) were determined using the recently developed molecular genetic algorithm. For compositions up to 18 atoms the algorithm found the global minima in all five runs. For clusters with less than 12 atoms, pure Au clusters are most stable. As the total number of atoms increases, clusters with one, two, three, or four Cu atoms become most stable. It has been observed for larger clusters that there is a tendency for structures based on Cu cores surrounded by Au atoms. The calculated second differences in binding energies are in fair agreement with experimental mass spectra for some nuclearities. Finally, the extrapolated cohesive energies for nanoalloys ((Cu3Au)(m), (CuAu3)(m), and (CuAu)(m)) gave an average relative error of 11%, compared with the bulk alloys, whereas for the nanoclusters (Au-N and Cu-N) the error was less than 6%. (C) 2003 Wiley Periodicals, Inc.