The slip distribution of the 12 November 1996, Mw = 7.7, Peru earthquake is determined using broadband teleseismic waveforms, a differential SAR interferogram (interferometric synthetic aperture radar [InSAR]), and a fault parametrization allowing slip and rupture velocity to vary along the rupture plane. Both data sets are inverted jointly to limit the trade‐off between the space and time aspects of the rupture. The earthquake fault plane is located at the subduction interface; it strikes parallel to the trench and dips 30° NE. By inverting synthetic data, we show how the InSAR and teleseismic data are complementary and how the joint inversion produces a gain in the spatial and temporal resolution of the slip model, even with a SAR interferogram that covers only part of the coseismic deformation. The rupture of the 1996 Peru event initiated on the southern flank of the subducted Nazca ridge and propagated unilaterally toward the southeast (along strike) for more than 100 km at a depth between 20 and 40 km. The area of maximum slip (6–7 m) is located 50 km southeast of the hypocenter. The total seismic moment is 4.4 × 10²⁰ N m (our joint inversion). The source time function is approximately 60 s long and presents three major pulses of moment release. The dominant one, which occurred between 30 and 45 s, does not correspond to the area of largest slip but to the rupture of a wide zone located about 100 km away from the hypocenter where slip reaches only 2–3 m. Computed coseismic coastal uplift correlates well with the location of raised marine terraces and with the topography of the coastal cordillera, suggesting that these features may be related to the repetition of 1996‐type events at the interface between the Nazca ridge and the South American plate.