Previous promising tests of the new M06 family of functionals in predicting ruthenium-metal phosphine bond dissociation energies (Zhao, Y.; Truhlar, D. G. Acc. Chem. Res. 2008, 41, 157) have been extended to a series of phosphine complexes of chromium, molybdenum, nickel, and ruthenium for which relevant experimental data are available. In addition to the M06 family of functionals, bond dissociation enthalpies have been calculated using a selection of density functionals and hybrid functionals based on the generalized gradient approximation (GGA), and with or without an empirical term (i.e., DFT-D) accounting for long-range dispersion. For the ruthenium complexes, second-order Møller-Plesset perturbation theory (MP2) has also been applied. Electrostatic and nonelectrostatic solvent effects have been estimated using the polarizable continuum model (PCM), allowing for comparison with experimental data obtained for dissociation reactions in organic solvents. Whereas the GGA and hybrid-GGA functionals grossly underestimate the absolute metal-phosphine bond enthalpies, with mean unsigned errors (MUEs) for a set of 10 phosphine dissociation reactions in the range 13-27 kcal/mol, the recently developed DFT-based methods for inclusion of attractive noncovalent interactions and dispersion (the DFT-D and M06 functionals) dramatically improve upon the situation. The best agreement with experiment is observed for BLYP-D (MUE = 2.2 kcal/mol), and with the exception for M06-2X, all these methods provide MUEs well below 5 kcal/mol, which should be sufficient for a broad range of applications. The improvements in predicted relative bond enthalpies are less convincing, however. In several cases the GGA and hybrid-GGA functionals are better at reproducing substitution effects than the DFT-D and M06 methods.