Description: The Atlantic meridional overturning circulation and meridional heat transport (hereafter the AMOC and MHT) at 34°S as simulated by global 1/16° eddy-rich (henceforth GLOB16) and ¼° eddy-permitting (GLOB4) models are compared with observational estimates. Three different methods are used for calculating the modelled AMOC: the first method (MOCmod) is based on simulated velocity fields, while the second (MOCob) relies on the same assumptions as available observed-derived estimates. The third method (MOCob2) is also based on hydrostatic and geostrophic relationships, but relative to a barotropic circulation instead of the definition of velocity at a specific reference depth. All methods correctly reproduce the time-mean GLOB16 AMOC strength, but the value of the non-Ekman component of the GLOB16 AMOC is only about 75% of the observed-derived estimate. The GLOB16 MHT is also significantly less than observation value (slightly more than 60% of the observed). However, the mean AMOC and MHT values at 34°S obtained with coarser resolution GLOB4 model are comparable with the observed-derived estimates. Possible causes for the differences between the eddy-rich model and observational data are studied. It is shown that the density field from the eddy-rich model has high temporal variability along 34°S with spatial scale of about two hundred km that can be due to mesoscale variations, caused by Agulhas “leakage”. This results in the decrease of the mean meridional geostrophic velocity, which leads to smaller values of the AMOC and MHT in GLOB16: subsampling GLOB16 density on ¼° or ½° longitude grid along 34°S for MOCob calculation significantly increases the AMOC values. The findings in this paper provide guidance in understanding AMOC and MHT dissimilarities between observation-based estimates and eddy-rich ocean models at 34°S. This article is protected by copyright. All rights reserved.