Two effective computational approaches for the study of magnetic exchange interactions in large molecules are discussed and tested on a number of model systems, namely, broken-symmetry (BS) and single-determinant (SD) models. Both methods are based on the density functional theory (DFT) but exploit different approximations to deal with multiconfigurational problems. Our results show that the BS model provides semiquantitative results for widely different situations, such as metal-radical interactions and metal-metal interactions mediated by inert organic bridges. Although more refined (and expensive) methods are needed for truely quantitative work, the BS/DFT approach provides a very useful tool for the rationalization of magneto-structural correlations and for the comparison of different bonding situations in large systems involving transition metal atoms.