Analyses and modeling of gravity data in the Dead Sea pull‐apart basin reveal the geometry of the basin and constrain models for its evolution. The basin is located within a valley which defines the Dead Sea transform plate boundary between Africa and Arabia. Three hundred kilometers of continuous marine gravity data, collected in a lake occupying the northern part of the basin, were integrated with land gravity data from Israel and Jordan to provide coverage to 30 km either side of the basin. Free‐air and variable‐density Bouguer anomaly maps, a horizontal first derivative map of the Bouguer anomaly, and gravity models of profiles across and along the basin were used with existing geological and geophysical information to infer the structure of the basin. The basin is a long (132 km), narrow (7–10 km), and deep (≤10 km) full graben which is bounded by subvertical faults along its long sides. The Bouguer anomaly along the axis of the basin decreases gradually from both the northern and southern ends, suggesting that the basin sags toward the center and is not bounded by faults at its narrow ends. The surface expression of the basin is wider at its center (<16 km) and covers the entire width of the transform valley due to the presence of shallower blocks that dip toward the basin. These blocks are interpreted to represent the widening of the basin by a passive collapse of the valley floor as the full graben deepened. The collapse was probably facilitated by movement along the normal faults that bound the transform valley. We present a model in which the geometry of the Dead Sea basin (i.e., full graben with relative along‐axis symmetry) may be controlled by stretching of the entire (brittle and ductile) crust along its long axis. There is no evidence for the participation of the upper mantle in the deformation of the basin, and the Moho is not significantly elevated. The basin is probably close to being isostatically uncompensated, and thermal effects related to stretching are expected to be minimal. The amount of crustal stretching calculated from this model is 21 km and the stretching factor is 1.19. If the rate of crustal stretching is similar to the rate of relative plate motion (6 mm/yr), the basin should be ∼3.5 m.y. old, in accord with geological evidence.