The GPS‐derived crustal velocity field of the western United States is used to construct dislocation models in a viscoelastic medium of interseismic crustal deformation. The interseismic velocity field is constrained by 1052 GPS velocity vectors spanning the ∼2500‐km‐long plate boundary zone adjacent to the San Andreas fault and Cascadia subduction zone and extending ∼1000 km into the plate interior. The GPS data set is compiled from U.S. Geological Survey campaign data, Plate Boundary Observatory data, and the Western U.S. Cordillera velocity field of Bennett et al. (1999). In the context of viscoelastic cycle models of postearthquake deformation, the interseismic velocity field is modeled with a combination of earthquake sources on ∼100 known faults plus broadly distributed sources. Models that best explain the observed interseismic velocity field include the contributions of viscoelastic relaxation from faulting near the major plate margins, viscoelastic relaxation from distributed faulting in the plate interior, as well as lateral variations in depth‐averaged rigidity in the elastic lithosphere. Resulting rigidity variations are consistent with reduced effective elastic plate thickness in a zone a few tens of kilometers wide surrounding the San Andreas fault (SAF) system. Primary deformation characteristics are captured along the entire SAF system, Eastern California Shear Zone, Walker Lane, the Mendocino triple junction, the Cascadia margin, and the plate interior up to ∼1000 km from the major plate boundaries.