The temperatures of internal components within spent nuclear fuel (SNF) packages must be predicted to ensure they remain within specified limits. These packages have multiple millimeter-scale gaps whose widths are affected by predictable and unpredictable factors. The thermal resistance across these gaps is proportional to their width and it is relatively large because the thermal conductivity of the gas within the packages is significantly lower than that of the metal components. Prior numerical research has demonstrated that variations in gap widths can result in peak cladding temperature varying by as much as 68 °C. In this work, a three-dimensional computational fluid dynamics model of the TN-32B SNF package that employs effective fuel region and gap properties is constructed. Forty steady-state simulations are performed for varying widths of nine different gaps within their possible ranges using the Latin Hypercube Sampling method. The best estimate of temperatures and their uncertainties at 63 SNF fuel locations are compared with measurements acquired from the High Burnup Spent Fuel Data project, where other researchers collected temperature data from a TN-32B SNF cask. The results showed that predicted temperatures have 95%-confidence-level interval uncertainties ranging between ±12 °C and ±21 °C due to uncertainties in the gap widths. Most of the measured temperatures, including the peak cladding temperature, are within the confidence intervals of the predicted ones. A proposed linear correlation between the simulated fuel temperatures and gap widths recreated 95% of the simulation results within ±1.6 °C. This correlation highlighted that uncertainty in the width of the Basket/Rail gap near the package periphery accounts for approximately half of the total temperature uncertainty.