The 1992 version of MM3 was largely used for modeling mono-, di-, and trisaccharides. In later versions of MM3 improvements were made in some parameters that may be important for carbohydrates. This corrected MM3 force field is part of the Tinker package, freely available (as its 4.1 version), and included in the Chem 3D Ultra 8.0 package (as the 3.7 version). The latter version lacks the corrections to the standard bond lengths produced by electronegativity and anomeric effects, whereas the Tinker 4.1 version only lacks the latter correction. The present work compares the performance of the three MM3 versions (and in some cases, DFT and/or HF/ab initio procedures) on several carbohydrate model problems as the chair and rotamer equilibria in 2-hydroxy- and 2-methoxytetrahydropyran, hydrogen bonding in cis-2,3-dihydroxytetrahydropyran, and the potential energy surfaces around the glycosidic bonds of two sulfated disaccharides and two trisaccharides. Tinker MM3 can be used accurately to estimate carbohydrate energies and geometries, and-with the help of some programming-to pursue studies on the potential energy surfaces of di- and trisaccharides. In most cases results obtained using the three MM3 versions are similar, although large energy differences are obtained when comparing a rotameric distribution around a O-C-O-H dihedral, which is almost forced to the exo-anomeric position by the Tinker versions. In other systems smaller energy differences are found, but they can nevertheless lead to a different global minimum when comparing conformers of similar energy. MM3(92) establishes better the differences between the bond lengths in both anomers, as an expected expression of the anomeric correction.