Global climate models (GCMs) have, in general, problems representing precipitation over southern southeastern South America (SSESA), namely southern Brazil, Uruguay and eastern Argentina. In this study, 18 models were selected from the Coupled Model Inter‐comparison Project Phase 5 to evaluate rainfall of the warm semester (October–March) when most of the rain takes place. Though GCMs reproduce the main features of the two main low‐level circulation variability modes over southeastern South America, east of the Andes, particularly the two flow patterns associated with the active and weak SACZ, most of them underestimate the northern flow that brings water vapour into SSESA, as well as the mean precipitation in this region. According to an index that stands for the northern flow towards SSESA, all models, except one, have lower northern flow into SSESA than the observed reanalysis field. In addition, this index has a significant correlation across GCMs with mean SSESA precipitation, indicating that lower northern flow into SSESA is generally associated with a greater underestimation in the model precipitation. Hence, it is concluded that the weaker northern flow into SSESA simulated by most GCMs is a cause of their simulated lower precipitation in this region.