Mantle rheology is one of the essential, yet least understood, material properties of our planet, controlling the dynamic processes inside the Earth's mantle and the Earth's response to various forces. With the advent of GRACE satellite gravity, measurements of mass displacements associated with many processes are now available. In the case of mass displacements related to postseismic deformation, these data may provide new constraints on the mantle rheology. We consider the postseismic deformation due to the M_w = 9.2 Sumatra 26 December 2004 and M_w = 8.7 Nias 28 March 2005 earthquakes. Applying wavelet analyses to enhance those local signals in the GRACE time varying geoids up to September 2007, we detect a clear postseismic gravity signal. We supplement these gravity variations with GPS measurements of postseismic crustal displacements to constrain postseismic relaxation processes throughout the upper mantle. The observed GPS displacements and gravity variations are well explained by a model of viscoelastic relaxation plus a small amount of afterslip at the downdip extension of the coseismically ruptured fault planes. Our model uses a 60 km thick elastic layer above a viscoelastic asthenosphere with Burgers body rheology. The mantle below depth 220 km has a Maxwell rheology. Assuming a low transient viscosity in the 60–220 km depth range, the GRACE data are best explained by a constant steady state viscosity throughout the ductile portion of the upper mantle (e.g., 60–660 km). This suggests that the localization of relatively low viscosity in the asthenosphere is chiefly in the transient viscosity rather than the steady state viscosity. We find a 8.10¹⁸ Pa s mantle viscosity in the 220–660 km depth range. This may indicate a transient response of the upper mantle to the high amount of stress released by the earthquakes. To fit the remaining misfit to the GRACE data, larger at the smaller spatial scales, cumulative afterslip of about 75 cm at depth should be added over the period spanned by the GRACE models. It produces only small crustal displacements. Our results confirm that satellite gravity data are an essential complement to ground geodetic and geophysical networks in order to understand the seismic cycle and the Earth's inner structure.