Thirty-five sets of thermal reactor data were analyzed mathematically to derive a best fit function to predict gas release from boron carbide as a function of temperature, irradiation exposure, and material density. An exposure variable to account for difference in self-shielding in various reactor spectra was developed. The data used in the analysis included temperatures from 550 to 1200°F, irradiation exposures from 10 × 1020 to 31 × 1020 captures/g, and material densities of 2.0 and 2.5 g/cm3 (80 and 99% of the theoretical density). Within this range the function should predict gas release in all reactor spectra within the 20% estimated accuracy of the experimental data. Independent gas release data generated in fast and intermediate reactor spectra showed that the general form of the function is correct but that application to temperature conditions above the specified range may result in overestimates of gas release.