δ13CO2 measured in Antarctic ice cores provides constraints on oceanic and terrestrial carbon cycle processes linked with millennial-scale changes in atmospheric CO2. However, the interpretation of δ13CO2 is not straightforward. Using carbon isotope-enabled versions of the LOVECLIM and Bern3D models, we perform a set of sensitivity experiments in which the formation rates of North Atlantic Deep Water (NADW), North Pacific Deep Water (NPDW), Antarctic Bottom Water (AABW) and Antarctic Intermediate Water (AAIW) are varied. We study the impact of these circulation changes on atmospheric δ13CO2 as well as on the oceanic δ13C distribution. In general, we find that the formation rates of AABW, NADW, NPDW and AAIW are negatively correlated with changes in δ13CO2: namely strong oceanic ventilation decreases atmospheric δ13CO2. However, since large scale ocean circulation reorganizations also impact nutrient utilization and the Earth's climate, the relationship between atmospheric δ13CO2 levels and ocean ventilation rate is not unequivocal. In both models atmospheric δ13CO2 is very sensitive to changes in AABW formation rates: increased AABW formation enhances the transport of low δ13C waters to the surface and decreases atmospheric δ13CO2. By contrast, the impact of NADW changes on atmospheric δ13CO2 is less robust and might be model dependent. This results from complex interplay between global climate, carbon cycle, and the formation rate of NADW, a water body characterized by relatively high δ13C.