Two-dimensional, plane-strain finite element numerical models produce small normal faults similar to those formed during the Oligocene to Miocene of the salt-cored St. Malo anticline in the deepwater Gulf of Mexico. The mechanical stratigraphy used in the models was derived from well data and a rate-independent, elastic-plastic constitutive model with a non-associated flow rule was used to represent the behavior of weak, over-consolidated rocks. Motion of salt is modeled by displacing the base of the overburden from an initially flat configuration to the observed present-day geometry. Model results using dominantly vertical displacement with minor extension (2%) are consistent with observed faulting at St. Malo. Small amounts of contraction (2–5%) in the numerical model suppress normal faulting whereas 2% extension best reproduces the observed structural style. The normal faults develop during elastic-plastic bending and evolve from sub-vertical, plastic mode-I failure zones to dominantly inclined normal faults. Throws of normal faults produced by the numerical models range from 11 m to 123 m. By comparison, the throws observed in the crest of the St. Malo anticline range from 30 m to 300 m. Models only using vertical displacement develop normal faults with ≤50 m throws due to bending; these are below seismic resolution. Only models with ≥2% extension develop normal faults that would be detectible (i.e., throws ≥50 m). A constraint of all the models is that the top of the salt is not faulted. The maximum depth of normal faulting in the models is ca. 900 m below the top of the reservoir. The maximum throws at the top of the reservoir in the models are ca. 30–65 m. Initiation of the normal faults as plastic mode-I failure zones in the numerical models suggests a mechanism that could facilitate early seal breach, even without juxtaposition of stratal leak points. •Numerical models reproduce normal faults in crest of salt-cored St. Malo anticline.•Calibration of model results with observations predicts subseismic faults.•Rock properties and finite element mesh size control strain localizations.