Previously, the US Nuclear Regulatory Commission performed simulations using fire dynamics simulator (FDS) software to predict the response of spent nuclear fuel transport packages in severe naturally ventilated tunnel fires. The long-range objective of the authors' current project is to predict the response of such a package to those tunnel fires using different computational methods. The first stage of the project, which is the subject of this article, is to determine the accuracy of Container Analysis Fire Environment (CAFE) computer simulations in predicting gas speed and temperature measurements made in forced-ventilated and naturally ventilated fires from the Memorial tunnel test series performed for the Massachusetts Highway Department in the 1990s. The CAFE simulations accurately predict the average heat release rate in both types of tests. Gas speeds and temperatures are obtained from the simulations at the same locations as for the measurements. For the forced-ventilated test, CAFE predicts those quantities more accurately upstream of the fire than downstream. In addition, the predictions are less accurate for the naturally ventilated test than they are for the forced-ventilated experiment. However, the accuracy of the CAFE prediction in the naturally ventilated test is on par with that from FDS.