On 29 July 2021, an Mw 8.2 megathrust earthquake struck the Alaska Peninsula. Quantifying the coseismic slip and the afterslip that followed this earthquake provides us the opportunity to clarify the megathrust slip budget and the earthquake hazard potential there. However, the estimated coseismic slip distribution inversion result is strongly affected by assumptions made in the inversion. The spatial pattern of stress-driven afterslip is mainly controlled by the coseismic slip distribution, so that it can provide new information about the coseismic slip distribution and is useful to assess the assumptions made in the coseismic inversion. The orientation and relative magnitudes of postseismic displacements at sites on the Alaska Peninsula require that the afterslip be concentrated ~130km from the trench. As a result, coseismic slip models including slip at that distance or less to shore, predict postseismic deformation that systematically misfits the observations. A narrower coseismic rupture plane with an abrupt downward termination of slip provides a much better fit to the observed postseismic signal than models where the slip tapers gently with depth. We considered multiple different viscoelastic relaxation models and find that these conclusions about the coseismic model are required regardless of the viscoelastic relaxation models used. The contribution of viscoelastic relaxation to the observed signal is not negligible, and the early postseismic observations are best reproduced with a model that features a 50 km thick elastic lithosphere for the overriding plate, and an elastic cold nose to the mantle wedge.