The assumption that abiotic air-sea gas exchange is, via the temperature dependence of the gas' solubility, proportional to the surface heat flux is often used to distinguish between physically and biotically inferred oxygen fluxes across the sea surface. We quantitatively investigate its validity in the context of an eddy-permitting circulation model that contains an abiotic oxygen compartment. In the model, the ``true'' abiotic oxygen air-sea fluxes are systematically lower than those predicted by the air-sea heat flux relation. This discrepancy is caused by the nonlinear relationship between temperature and solubility that results in the saturation of a mixed water parcel being higher than the arithmetic mean saturation of the mixed components. This effect results in a simulated additional sea-to-air oxygen flux of about 0.5 mol O2 m-2 a-1 north of 40°N, which is not accounted for by the heat-flux relation and which is of similar magnitude as, though at the lower end of, biotically induced oxygen fluxes. Simulated outgassing of the model's abiotic oxygen is also higher than that predicted by the heat-flux relation at the equator (by ~0.25 mol O2 m-2 a-1), where numerical artifacts endemic to state-of-the-art z level ocean models are found to affect simulated air-sea gas exchange. In addition to discrepancies in the annual mean fluxes, model results also indicate that the subtropical seasonal cycle in abiotic air-sea oxygen exchange is smaller by approximately 20% than the estimate based on air-sea heat fluxes, a result consistent with admittedly sparse observations of argon saturation.