The predictive capabilities of blade-resolved unsteady Reynolds averaged Navier-Stokes (URANS) and detached eddy simulation (DES), the most commonly used hybrid RANS/large eddy simulation (LES) model, are assessed for hydrokinetic turbine performance and mean and turbulent flows in the intermediate-wake region, and results for a range of tip-speed ratio encompassing design and off-design conditions are analyzed to understand the wake recovery mechanism. The performance predictions compared within 5% of the experimental data. Both URANS and DES models performed reasonably well for the near wake predictions, where the errors were < 15%. DES outperformed URANS for both mean wake deficit and turbulence predictions in the intermediate-wake region and both quantities compared within 10% of the experiments. The improved prediction by DES is because of 1) its ability to predict the tip vortex breakdown, which plays a critical role in the wake recovery, especially for higher tip speed ratios; 2) the presence of the free-surface which created an upper bypass region of accelerated flow. The study reveals that the tip vortex breakdown mechanism depends on tip speed ratio. For lower values of tip speed ratio, instabilities generated in the root vortex core are identified to be the cause of breakdown. For higher values, the breakdown occurred because of the instabilities generated during the vortex filament entanglement. The presence of the free-surface led to an early vortex breakdown and the interaction between the wake and free-surface is initiated by the interaction of stanchion with the free-surface. Future work should focus on investigation of other hybrid RANS/LES models to address the limitations of the DES models, and extension of the study to include wave effects.