Observational evidence from deep‐sea cores suggests that the ocean circulation during the last glacial cycle was highly variable, at times occupying states very different from those found today. Modelling can be used to dynamically constrain the possible circulation states compatible with observations, thus guiding both understanding of past climate but also the geographical and scientific thrust of future palaeoceanographic research. The Last Glacial Maximum has been extensively studied and here, using carbon isotopes, among other variables, the most likely thermohaline state consistent with palaeoclimatic data is constructed. Past and new modelling efforts for the Last Glacial Maximum are then examined, to contrast questions resolvable by the modelling/data comparison with those that remain unanswered. This shows modelling evidence to confirm the prevailing view of intermediate‐depth North Atlantic Deep Water being produced at the Last Glacial Maximum, in combination with deep water production around Antarctica. The potential sites for this deep and intermediate water production are defined by the basic state of the thermohaline circulation. However, their relative importance is a function of small perturbations in the surface temperature and salinity fields brought about by active coupling between the ocean and atmosphere. Regions where these water masses may have been produced at the Last Glacial Maximum are suggested.