A different number of bridging carbonyls is found in bi- or trinuclear clusters having the title formulas. Comparative calculations at the SCF, MP2, and DFT levels of theory show that only the latter is able to describe properly the energetics of various isomers of the whole triad. For the first-row transition metal, DFT gives excellent agreement with the experimental structures, whereas the MP2 approach fails completely. Conversely for the second- and third-row metals, the best agreement with the experiment is obtained by the MP2 optimizations. The quantitative computational results, associated with a qualitative MO analysis, allow one to conclude that the structural preferences are determined by a critical balance of metal-bridge bonding, metal-metal bonding, and intermetallic repulsion. Although the M-M bond order is expected to be 1 in all cases, the bridge-supported bond is experimentally and computationally shorter than the unsupported one. By contrast, the trend for the overlap population (OP) is reversed, with even negative Values for the shorter bridge bonds. For the latter, only a weak attractive interaction stems from the almost pure t(2g) orbitals, taken as metal lone pairs or eventually responsible for back-donation (formation of metal-bridge sigma bonds). Thus, the negative OP values are consistent with a prevailing repulsion between the latter levels. In the iron systems, with more contracted metal orbitals, the direct metal-metal repulsion is relatively weak while the metal-bridge bonds are sufficiently strong. This is not equally true for the more diffuse ruthenium and osmium orbitals, so the alternative nonbridged structure is preferred.