We use about two decades of geodetic measurements to characterize interseismic strain build up along the Central Andes subduction zone from Lima, Peru, to Antofagasta, Chile. These measurements are modeled assuming a 3-plate model (Nazca, Andean sliver and South America Craton) and spatially varying interseismic coupling (ISC) on the Nazca megathrust interface. We also determine slip models of the 1996 Mw = 7.7 Nazca, the 2001 Mw = 8.4 Arequipa, the 2007 Mw = 8.0 Pisco and the Mw = 7.7 Tocopilla earthquakes. We find that the data require a highly heterogeneous ISC pattern and that, overall, areas with large seismic slip coincide with areas which remain locked in the interseismic period (with high ISC). Offshore Lima where the ISC is high, a Mw∼8.6-8.8 earthquake occurred in 1746. This area ruptured again in a sequence of four Mw∼8.0 earthquakes in 1940, 1966, 1974 and 2007 but these events released only a small fraction of the elastic strain which has built up since 1746 so that enough elastic strain might be available there to generate a Mw > 8.5 earthquake. The region where the Nazca ridge subducts appears to be mostly creeping aseismically in the interseismic period (low ISC) and seems to act as a permanent barrier as no large earthquake ruptured through it in the last 500years. In southern Peru, ISC is relatively high and the deficit of moment accumulated since the M w∼8.8 earthquake of 1868 is equivalent to a magnitude M w∼8.4 earthquake. Two asperities separated by a subtle aseismic creeping patch are revealed there. This aseismic patch may arrest some rupture as happened during the 2001 Arequipa earthquake, but the larger earthquakes of 1604 and 1868 were able to rupture through it. In northern Chile, ISC is very high and the rupture of the 2007 Tocopilla earthquake has released only 4% of the elastic strain that has accumulated since 1877. The deficit of moment which has accumulated there is equivalent to a magnitude Mw∼8.7 earthquake. This study thus provides elements to assess the location, size and magnitude of future large megathurst earthquakes in the Central Andes subduction zone. Caveats of this study are that interseismic strain of the forearc is assumed time invariant and entirely elastic. Also a major source of uncertainty is due to fact that the available data place very little constraints on interseismic coupling at shallow depth near the trench, except offshore Lima where sea bottom geodetic measurements have been collected suggesting strong coupling.