Ocean heat transport in the South Atlantic of a coupled climate model has been diagnosed and compared with observational estimates. The coupled model overestimates the northward heat transport in the South Atlantic. This corresponds to an underestimate of the northward flux of bottom water across World Ocean Circulation Experiment (WOCE) section A11 (the model transports less than 1 Sv compared with climatological estimates of 6 Sv). The magnitude of the southward outflow of North Atlantic deep water agrees well with observational estimates while the northward flux of surface and intermediate waters is larger than observational estimates. We show that with the correct water mass properties a realistic northward flux of bottom water across 30°S is only possible, in the climate model, if the depths of the channels linking the Argentine and Brazil basins are properly represented. We find that an increase in the northward flow of bottom water corresponds to an increase in the southward flow of deep water, rather than a reduction in the northward flux of surface and intermediate waters. This indicates that the upper and lower limb of the overturning are decoupled in the model. As a consequence, increasing the northward transport of bottom water leads to a small reduction of the northward heat transport across 30°S. The magnitude of the heat transport from the Indian to the Atlantic ocean in the model (the warm water path) is surprisingly large for a noneddy resolving model. We find that the magnitude of the Indonesian Throughflow is important in climate resolution models for determining the strength of the warm water path. However, the relative strengths of the warm and cold water paths do not significantly change the heat transported across 30°S.