The ability of a dinucleoside monophosphate to mimic the conformation of B-DNA was tested using a combination of different phosphate models (anionic, neutral, counterion), environments (gas, water), and density functionals (B3LYP, MPWB1K, M06-2X) with the 6-31G(d,p) basis set. Three sequences (5′-GXPy-3′, where XPy = T, U or BrU) were considered, which vary in the (natural or modified) 3′ pyrimidine nucleobase (XPy). These bases were selected due to their presence in natural DNA, structural similarity to T and/or applications in radical-initiated anti-tumour therapies. The accuracy of each of the 54 model, method and sequence combinations was judged based on the ability to reproduce key structural features of natural B-DNA, including the stacked base–base orientation and important backbone torsion angles. B3LYP yields distorted or tilted relative base–base orientations, while many computational challenges were encountered for MPWB1K. Despite wide use in computational studies of DNA, the neutral (protonated) phosphate model could not consistently predict a stacked arrangement for all sequences. In contrast, stacked base–base arrangements were obtained for all sequences with M06-2X in conjunction with both the anionic and (sodium) counterion phosphate models. However, comparison of calculated and experimental backbone conformations reveals the charge-neutralized counterion phosphate model best mimics B-DNA. Structures optimized with implicit solvent (water) are comparable to gas-phase structures, which suggests similar results should be obtained in an environment of intermediate polarity. We recommend the use of either gas-phase or water M06-2X optimizations with the counterion phosphate model to study the structure and/or reactivity of natural or modified DNA with a dinucleoside monophosphate.