Atmospheric measurements of O2/N2 and CO2 at up to nine sites have been used to infer the interannual variations in oceanic O2 exchange with an inverse method. The method distinguishes the regional contributions of three latitudinal bands, partly the individual contributions of the North Pacific and the North Atlantic also. The interannual variations of the inferred O2 fluxes in the tropical band correlate significantly with the El Niño/Southern Oscillation. Tropical O2 variations appear to be dominated by the ventilation of the O2 minimum zone from variations in Pacific equatorial upwelling. The interannual variations of the northern and southern extratropical bands are of similar amplitude, though the attribution to mechanisms is less clear. The interannual variations estimated by the inverse method are larger than those estimated by the current generation of global ocean biogeochemistry models, especially in the North Atlantic, suggesting that the representation of biological processes plays a role. The comparison further suggests that O2 variability is a more stringent test to validate models than CO2 variability, because the processes driving O2 variability combine in the same direction and amplify the underlying climatic signal.