The effective elastic thickness of the lithosphere Te has been investigated by inverse and forward models based on the assumption that flexure of the surface and crust-mantle boundaries produces gravity field variations which are commonly expressed in free air or Bouguer gravity anomalies. The aim of this study is to examine Te variations across the East African and Baikal rifts that have developed within or adjacent to early Proterozoic-Archean cratons in order to investigate the possibility that the crust and upper mantle beneath the cratons retain considerable strength in extension. We use forward models of five 1800-km-long Bouguer gravity profiles in the East African rift and one profile previously modeled in the Baikal rift. The forward modeling method uses a multilayered lithospheric rheology to compute the mechanical plate deflection in response to vertical loads. We show that the origin of the rift topography may be highly variable: uplifted rift flanks are few in the Baikal rift, where mountain ranges have generally deep crustal roots, while flexural rebound accounts for most of the short-wavelength topography encountered in the East African rift. Reasonable fits of models and observed Bouguer gravity anomalies can only be obtained if we allow large rheological contrasts between the cratons and surrounding areas; Te values reach 60 km beneath the cratons and range between 20 and 45 km beneath the surrounding orogenic belts. The major part of the rifts develops close to the craton margins, where rheological contrasts may cause stress concentrations. These results agree well with previous Bouguer coherence analyses but are significantly different from previous free-air coherence models and forward models of topography alone, which yielded Te estimates <10 km beneath the rifts and the surrounding areas. We believe that these differences largely stem from model assumptions of compensation depth and lithospheric mechanical properties and from the fact that free-air anomalies contain short-wavelength noise that correlates with topography, which biases free-air based methods toward low Te estimates. ¿ 2000 American Geophysical Union