The flow field of a nonpremixed industrial gas burner is analyzed with Reynolds-averaged Navier–Stokes computational fluid dynamics validated against velocity and pressure measurements. Combustion is not modeled because the aim is optimizing the predictive capabilities of the cold flow before including chemistry. The system's complex flow physics, affected by a 90 deg turn, backward and forward facing steps, and transversal jets in the mainstream is investigated at full and partial load. The sensitivity of the computed flow field to inflow boundary condition setup, approach for resolving/modeling wall-bounded flows, and turbulence closure is assessed. In the first sensitivity analysis, the inflow boundary condition is prescribed using measured total pressure or measured velocity field. In the second, boundary layers are resolved down to the wall or modeled with wall functions. In the third sensitivity analysis, the turbulence closure uses the shear stress transport eddy viscosity model or two variants of the Reynolds stress model. The agreement between the predictions of most simulation setups among themselves and with the measurements is good. For a given type of inflow condition and wall flow treatment, the ω-based Reynolds stress model gives the best agreement with measurements among the considered turbulence models at full load. At partial load, the comparison with measured data highlights some scatter in the predictions of different patterns of the flow measurements. Overall, the findings of this study provide insight into the fluid dynamics of industrial gas burners and guidelines for their simulation-based analysis.