A computational model is described for fission product release from molten pools of uranium-aluminum (U-Al) metal. Liquid-metal pools may form during severe accidents in U-Al-fueled reactors if multiple core assemblies melt and relocate to the bottom of the reactor vessel. At present, data for the release of fission products from intact U-Al fuel are sparse, and no data are available for the release of fission products from U-Al in the form of molten pools. This investigation postulates three phenomena that govern fission product release from such a system: (a) Rayleigh cell convection in the molten pool; (b) nucleation of volatile radionuclide species with concomitant bubble dynamics; and (c) diffusion from the pool surface. Selected sensitivity analyses have been performed to study the dependence of model predictions on uncertain input parameters and thus to characterize critical needs for experimental data. The results of the sensitivity analysis indicate that parameters that characterize the nucleation rate of volatile species in the pool have the greatest effect on the calculated rate of fission product release.