A rigorous gas-liquid mass transfer model based on Maxwell-Stefan diffusion and two-film theory was validated against dynamic oxygen transfer experiments and exit gas analysis with mass spectrometry. A multiblock stirred tank model, which consists of 21 ideally mixed subregions, was used to consider non-ideal mixing in the model validation. Mass transfer areas were obtained from population balances for bubbles with experimentally validated breakage and coalescence closures. The model was implemented to a commercial computational fluid dynamic (CFD) program to investigate local mass transfer in a 200 L laboratory tank. The computational cost of CFD was more than 1000-fold compared to the multiblock model, but the results are comparable. The multiblock and CFD simulations show strongly inhomogeneous mass transfer in the agitated tank. This mostly results from spatially varying gas-liquid interfacial areas but also from varying gas phase concentrations and turbulence energy dissipations. A comparison to the semi-empirical k(L)a-correlations shows that non-ideal mixing must be considered in the analysis of mass transfer experiments to avoid the dependence of model on flow field. Therefore, k(L)a-correlations may not be applicable to the investigation of local mass transfer in varying vessel geometries.