A series of medium-resolution (∼1°) numerical simulations for the equatorial and North Atlantic basin have been performed with two primitive equation models, one employing depth and the other density as the vertical coordinate. The models have been configured for this exercise in as similar a fashion as their basic formulations allow, and with fundamentally identical initialization, boundary conditions, and forcing functions for each of the experiments. The purpose of comparing the models’ results is twofold: 1) to understand the degree to which model-generated circulation fields depend on the particular model architecture by examining the rate of divergence of the solutions of an isopycnic and a depth coordinate model given the same initial conditions and 2) to uncover and remedy possible defects in either model design. The comparison is focused on the importance in each simulation of the choice of mixing parameterization, which has a crucial impact on the meridional overturning circulation, on the associated northward heat transport, and on the evolution of water masses. Although the model-generated horizontal fields viewed at specific times during the integrations do not appear to be strongly dependent on the design of each model and are in good agreement with one another, the integrated properties of the depth coordinate model and the isopycnic coordinate model diverge significantly over time, with the depth coordinate model being unable to retain its most dense water masses after long integration periods.