Completely ab initio global potential energy surfaces (PESs) for the singlet and triplet spin multiplicities of rigid O(2)((3)Σ(g)(-))+O(2)((3)Σ(g)(-)) are reported for the first time. They have been obtained by combining an accurate restricted coupled cluster theory with singles, doubles, and perturbative triple excitations [RCCSD(T)] quintet potential [Bartolomei et al., J. Chem. Phys. 128, 214304 (2008)] with complete active space second order perturbation theory (CASPT2) or, alternatively, multireference configuration interaction (MRCI) calculations of the singlet-quintet and triplet-quintet splittings. Spherical harmonic expansions, containing a large number of terms due to the high anisotropy of the interaction, have been built from the ab initio data. The radial coefficients of these expansions are matched at long range distances with analytical functions based on recent ab initio calculations of the electric properties of the monomers [M. Bartolomei, E. Carmona-Novillo, M. I. Hernández, J. Campos-Martínez, and R. Hernández-Lamoneda, J. Comput. Chem. (2010) (in press)]. The singlet and triplet PESs obtained from either RCCSD(T)-CASPT2 or RCCSD(T)-MRCI calculations are quite similar, although quantitative differences appear in specific terms of the expansion. CASPT2 calculations are the ones giving rise to larger splittings and more attractive interactions, particularly in the region of the absolute minima (in the rectangular D(2h) geometry). The new singlet, triplet, and quintet PESs are tested against second virial coefficient B(T) data and, their spherically averaged components, against integral cross sections measured with rotationally hot effusive beams. Both types of multiconfigurational approaches provide quite similar results, which, in turn, are in good agreement with the measurements. It is found that discrepancies with the experiments could be removed if the PESs were slightly more attractive. In this regard, the most attractive RCCSD(T)-CASPT2 PESs perform slightly better than the RCCSD(T)-MRCI counterpart.