We have determined the three-dimensional P-wave velocity structure beneath the north–south tectonic belt between Tibet and Eastern China by simultaneously inverting local, regional, and teleseismic data. Our data set is composed of 45,028 P-wave arrival times from 602 local and regional earthquakes and 985 travel times from 102 teleseismic events. With the lsqr (sparse linear equations and least squares) algorithm, the P-wave velocity perturbations were estimated by the simultaneous inversion of hypocenters and medium parameters from the surface to a depth of 200 km. We tested the stability and the resolution of our inverted results with a checkerboard test and found that the models are well resolved up to a depth of about 50 km for most parts of the studied region. Results show that the north–south tectonic belt is characterized by a significant lateral heterogeneity in velocity both in the crust and in the upper mantle. Correlating these velocity images with the main tectonic features, we find that (1) the shallow velocity distribution above ∼3 km is consistent with the topographic features and the basin distribution; (2) the middle-lower crustal velocities from 20 km to 50 km characterize a mechanically weak north–south tectonic belt, because it bears a relatively lower-velocity perturbation over a large region; (3) the upper mantle velocities from 85 km to 120 km delineate the eastern Tibetan boundary, but changes in some subzones may reflect the effects of several tectonic events, including paleorifting, the Cenozoic convergence between Tibet and southeastern China and other tectonic episodes.