In models of the global carbon cycle, the pCO2 of the atmosphere is more sensitive to the chemistry of the high-latitude surface ocean than the tropical ocean. Because sea-surface nutrient concentrations are generally high in the high latitudes, pCO2 sensitivity to high-latitude forcing also determines pCO2 sensitivity to the biological pump globally. We diagnose high-latitude sensitivity of a range of ocean models using atmospheric pCO2 above an abiotic ocean; cold high-latitude waters pull abiotic pCO2 to low values. Box models are very high-latitude sensitive, while most global circulation models are considerably less so, including a two-dimensional overturning model, two primitive equation models, the Hamburg class of large scale geostrophic (LSG) general circulation models (GCMs), and the MICOM isopycnic GCM. High-latitude forcing becomes more important in a depth-coordinate GCM when lateral diffusion is oriented along isopycnal surfaces, rather than horizontally, following Redi [1982]. In two different GCMs (a primitive equation model and LSG), addition of the Gent and McWilliams [1990] isopycnal thickness diffusion scheme had only minor impact on high-latitude sensitivity. Using a simplified box model, we show that high-latitude sensitivity depends on a high-latitude monopoly on deep water formation. In an attempt to bridge the gap between box models and GCMs, we constructed a simple slab overturning model with an imposed stream function which can be discretized at arbitrary resolution from box model to GCM scale. High-latitude sensitivity is independent of model resolution but very sensitive to vertical diffusion. Diffusion acts to break the high-latitude monopoly, decreasing high-latitude sensitivity. In the isopycnal GCM MICOM, however, high-latitude sensitivity is relatively insensitive to diapycnal diffusion of tracers such as CO2. This would imply that flow pathways in MICOM take the place of vertical diffusion in the slab model. The two nominally most sophisticated ocean models in the comparison are the isopycnal model MICOM and the depth-coordinate GCM with Redi [1982] and Gent and McWilliams [1990] mixing. Unfortunately, these two models disagree in their abiotic CO2 behavior; the depth-coordinate isopycnal mixing GCM is high-latitude sensitive, in accord with box models, while MICOM is less so. The rest of the GCMs, which have historically seen the most use in geochemical studies, are even less high-latitude sensitive than MICOM. This discrepancy needs to be resolved. In the meantime, the implication of the MICOM/traditional GCM result would be that box models overestimate high-latitude sensitivity of the real ocean. This would eliminate iron dust fertilization of the ocean as an explanation for the glacial pCO2 range of 180--200 &mgr;atm [Archer et al., 2000]. ¿ 2000 American Geophysical Union