The structure of B₈F₁₂ has been shown by gas electron diffraction and computational methods (up to MP2/6-31+G*) to have the same highly asymmetric form observed in crystalline phases. The structure can be regarded as derived from a central B₂ group, bridged by two BF₂ groups to give a central B₄ core that is folded, not planar, and with a very short bond [164.3 pm calculated, 164.2(19) pm experimental] along the fold line. There are also four terminal BF₂ groups. One of the other four bonds in the core is consistently 20–30 pm longer than the others. This asymmetry has been attributed to many intra-molecular B⋯F interactions, particularly those between core boron atoms and fluorines of the terminal BF₂ groups. Calculations for the chloro analogue lead to a structure similar to that for B₈F₁₂, but with the long core bond extended so that one of the bridging BCl₂ groups may now be regarded as terminal. With bromine as the halogen the structure changes again, with one bromine atom taking up a bridging position. With iodine, this process continues further, and there are three bridging iodine atoms. However, in this case this is not the lowest energy structure, and instead a loosely associated dimer of B₄I₆ is preferred. In all these cases, and particularly with the heavier halogens, there are huge differences between the results obtained with different computational methods.