The US Nuclear Regulatory Commission (NRC) participated in an Organization for Economic Cooperation and Development / Nuclear Energy Agency (OECD/NEA) benchmark activity based on testing in the PANDA facility located at the Paul Scherrer Institute in Switzerland. In this test, a stratified helium layer was eroded by a turbulent jet from below. NRC participated in this benchmark to develop expertise and modeling guidelines for computational fluid dynamics (CFD) in anticipation of utilizing these methods for future safety and confirmatory analyses. CFD predictions using ANSYS FLUENT V19.0 are benchmarked using the PANDA test data, and sensitivity studies are used to evaluate the significance of key phenomena, such as boundary conditions and modeling options, that impact the helium erosion rates and jet velocity distribution. The k-epsilon realizable approach with second order differencing resulted in the best prediction of the test data. The most significant phenomena are found to be the inlet mass flowrate and turbulent Schmidt number. CFD uncertainty for helium and velocity due to numerical error and input parameter uncertainty are predicted using a sensitivity coefficient approach. Numerical uncertainty resulting from the mesh design is estimated using a Grid Convergence Index (GCI) approach. Meshes of 0.5, 1.5 (base mesh), and 4.5 million cells are used for the GCI. Approximately second order grid convergence was observed but p (order of convergence) values from 1 to 5 were common. The final helium predictions with one-sigma uncertainty interval generally bounded the experimental data. The predicted jet centerline velocity was approximately 50% of the measured value at multiple measurement locations. This velocity benchmark is likely affected by the difference in the He content between the experiment and prediction. The predicted jet centerline velocity with the one-sigma uncertainty interval did not bound the experimental data.