Three different, eddy-permitting numerical models are used to examine the seasonal variation ofmeridional mass and heat flux in the North Atlantic, with a focus on the transport mechanisms inthe subtropics relating to observational studies near 25°N. The models, developed in theDYNAMO project, cover the same horizontal domain, with a locally isotropic grid of 1/3° resolutionin longitude, and are subject to the same monthly-mean atmospheric forcing based on athree-year ECMWF climatology. The models differ in the vertical-coordinate scheme(geopotential, isopycnic, and sigma), implying differences in lateral and diapycnic mixingconcepts, and implementation of bottom topography. As shown in the companion paper ofWillebrand et al. (2001), the model solutions exhibit significant discrepancies in the annual-meanpatterns of meridional mass and heat transport, as well as in the structure of the western boundarycurrent system.Despite these differences in the mean properties, the seasonal anomalies of the meridional fluxesare in remarkable agreement, demonstrating a robust model behavior that is primarily dependenton the external forcing, and independent of choices of numerics and parameterization. The annualrange is smaller than in previous model studies in which wind stress climatologies based on marineobservations were used, both in the equatorial Atlantic (1.4 PW) and in the subtropics (0.4-0.5PW). This is a consequence of a weaker seasonal variation in the zonal wind stresses based onthe ECMWF analysis than those derived from climatologies of marine observations.The similarities in the amplitude and patterns of the meridional transport anomalies betwen thedifferent model realizations provide support for previous model conclusions concerning themechanism of seasonal and intraseasonal heat flux variations: they can be rationalized in terms of atime-varying Ekman transport and their predominantly barotropic compensation at depth. Analysisfor 25°N indicates that the net meridional flow variation at depth is concentrated near the westernboundary, but cannot be inferred from transport measurements in the western boundary currentsystem, because of significant and complex recirculations over the western half of the basin. Themodel results instead suggest that the main requirement for estimating the annual cycle of heatflux through a transoceanic section, and the major source of error in model simulations, is anaccurate knowledge of the wind stress variation.