In this paper we use a biological-physical model with an explicit representation of Trichodesmium to examine the influence of N2 fixation in the Atlantic. Three solutions are examined, one where the N2 fixation rate has been set to observed levels, one where the rate has been increased to levels comparable to geochemical estimates, and one with no N2 fixation. All solutions are tuned to reproduce satellite surface chlorophyll concentrations, so that differences in the runs are manifested in productivity and export. Model runs with N2 fixation have different phytoplankton production and export distributions than runs without. Over the Atlantic basin the ecosystem fixes nitrogen at the rate of 1.47 ¿ 1012 mol N yr-1, when tuned to observed phytoplankton and Trichodesmium biomass. This rate is comparable to the lower range of direct estimates of 1.3--2.2 ¿ 1012mol N yr-1 <Capone et al., 1997; J. N. Galloway et al., manuscript in preparation, 2003; D. Capone et al., New nitrogen input in the tropical North Atlantic Ocean by nitrogen fixation, submitted to Nature, 2004, hereinafter referred to as Capone et al., submitted manuscript, 2004> but less than geochemical indirect estimates over a reduced domain (2.0 ¿ 1012 mol N yr-1 <Gruber and Sarmiento, 1997> versus 0.55 ¿ 1012 mol N yr-1 for the model). The nitrogen from N2 fixation increases new production by 30% and total production by 5%. However, it does not supplement upwelled nitrate sufficiently to bring production and export into line with remote sensing and geochemically derived estimates. Simulations with N2 fixation rates comparable to geochemical estimates show that reasonable phytoplankton concentrations can be maintained if export is increased. Moreover, phytoplankton productivity increases to values approaching remote-sensing-based estimates in the oligotrophic ocean. However, Trichodesmium biomass may be higher than observed.