The 2011 Tohoku-Oki earthquake with 9.0 Mw led to an enormous mass redistribution originated from large deformation due to faulting and had a massive impact on the coastal area of eastern Japan. While the satellite gravity mission GRACE (Gravity Recovery and Climate Experiment) can detect the gravitational change caused by this tremendous event, slip distributions are usually derived from GPS, seismic and (in the more particular case) tsunami data. We evaluate the differences between measured and modeled coseismic gravity changes for three fault slip models derived from either GPS and tsunami data, GRACE data, or a combination of all three data types. The data are weighted according to their measurement accuracy in a Bayesian joint inversion approach. We perform a long term average of GRACE data, which increases sensitivity and reduces artefacts, and find that the postseismic gravity change leaks into the derived mean gravity field. We try to reduce this problem by averaging only 6 months of postseismic GRACE data, where the postseismic gravity signal, which superimposes onto the coseismic signal of ≈ 6μGal (for a geometric based model) peaks approximately 3 months after earthquake occurrence. Consequently fault slip models merely derived from GPS (10 days avg.) and tsunami data (<5h time span) show deviations of ≈ 2 μGal to a GRACE 6 monthly averaged combined solution which indicates the difference accumulated from the geometric and gravimetric modeling and the postseismic gravity signal in the GRACE data.