Errors in global ocean models have been shown to be larger than the accuracy of surface gravity measurements <Baker and Bos, 2001> and also new space gravity missions <Ray et al., 2001>. In this paper, we compare the tidal loading estimated using 10 different ocean tide models to the surface gravity variations observed with 12 superconducting gravimeters (SG) belonging to the Global Geodynamics Project <Crossley et al., 1999> worldwide network. Ocean tidal gravity variations are shown to be mostly dependent on tidal sea height variations within a radius of ~5000 km. Precise surface gravity measurements are therefore another tool for validating ocean tidal models from short to long wavelengths in addition, for example, to tide gauge data. In this study we also compute the tidal loading using several different available computer codes. As for ocean tidal models, we show that the differences between tidal loading programs are frequently larger than the error estimates of SG measurements. Compared to previous models, both recent hydrodynamical and TOPEX-Poseidon derived tidal models allow a significant reduction of discrepancies between ocean tidal observations using SG and tidal loading estimates. There still remains, however, a nonnegligible in-phase discrepancy between ocean tidal loading estimates and surface gravity observations, whereas the out-of-phase component is much smaller. We have shown that the in-phase disagreement does not seem to be caused by an inaccurate determination of the SG calibration.