We use two gravity profiles that we measured across Central Nepal, in conjunction with existing data, to constrain the mechanical behaviour and the petrological structure of the lithosphere in the Himalayan collision zone. The data show (1) overcompensation of the foreland and undercompensation of the Higher Himalaya, as expected from the flexural support of the range; (2) a steep gravity gradient of the order of 1.3 mgalkm^-1 beneath the Higher Himalaya, suggesting a locally steeper Moho; and (3) a 10km wide hinge in southern Tibet. We compare these data with a 2-D mechanical model in which the Indian lithosphere is flexed down by the advancing front of the range and sedimentation in the foreland. The model assumes brittle Coulomb failure and non-linear ductile flow that depends on local temperature, which is computed from a steady-state thermal model. The computed Moho fits seismological constraints and is consistent with the main trends in the observed Bouguer anomaly. It predicts an equivalent elastic thickness of 40-50km in the foreland. The flexural rigidity decreases northwards due to thermal and flexural weakening, resulting in a steeper Moho dip beneath the high range. Residuals at short wavelengths (over distances of 20-30km) are interpreted in terms of (1) sediment compaction in the foreland (Δρ=150kgm^-3 between the Lower and Middle Siwaliks); (2) the contact between the Tertiary molasse and the meta-sediments of the Lesser Himalaya at the MBT (Δρ=220kgm^-3) and (3) the Palung granite intrusion in the Lesser Himalaya (Δρ=80kgm^-3). Finally, if petrological transformations expected from the local (P,T) are assumed, a gravity signature of the order of 250 mgal is predicted north of the Lesser Himalaya, essentially due to eclogitization of the lower crust, which is inconsistent with the gravity data. We conclude that eclogitization of the Indian crust does not take place as expected from a steady-state local equilibrium assumption. We show, however, that eclogitization might actually occur beneath southern Tibet, where it could explain the hinge observed in the gravity data. We suspect that these eclogites are subducted with the Indian lithosphere.