There have been no historical Cascadia great subduction thrust earthquakes, but there is good recent evidence that very large earthquakes have occurred in the past and that strain is building up toward a future great event. Geodetic measurements in the coastal region from northern California to southern British Columbia show vertical and horizontal deformation as expected for the strain accumulation of a locked thrust fault. The segment of the subduction thrust that is locked and may rupture in future great events has previously been estimated through two-dimensional (2-D) elastic dislocation modeling of interseismic deformation geodetic data. In this study, a general 3-D dislocation model for thrust faults has been developed that accommodates curved fault geometry and nonuniform interseismic locking or coseismic rupture. The model is based on the surface deformation due to shear faulting in an elastic half space. The 3-D model of the Cascadia subduction zone calculates the surface deformation for a locked zone or a rupture zone of variable width along the margin. The bend in the margin trend and subducting slab end effects are included. There is a downdip transition zone between interseismic completely locked and free slip portions of the fault or between coseismic full rupture and no displacement. An initial 3-D model based upon 2-D dislocation models and upon thermal constraints was adjusted to optimize the fit of the predicted interseismic surface deformation to current deformation geodetic data. The best fit model has the thrust locked along the whole margin with an average locked zone width of 60 km and a transition zone width of 60 km. The two zones lie mainly offshore beneath the continental shelf and slope. The locked and transition zone widths vary smoothly along the margin, being greater off northern Washington where the thrust dip is shallow and narrower off central Oregon. Assuming that the locked plus transition zones approximate the maximum coseismic rupture area, these widths permit a Mw=9 earthquake.¿ 1997 American Geophysical Union