Bentonite clay is considered as a potential buffer and backfill material in subsurface repositories for high-level nuclear waste. As a result of its low permeability, transport of water and solutes in compacted bentonite is driven primarily by diffusion. Developing models for species transport in bentonite is complicated, because of the interaction of charged species and the negative surface charge of clay mineral surfaces. The effective diffusion coefficient of an ion in bentonite depends on the ion's polarity and valence, on the ionic strength of the solution, and on the bulk dry density of the bentonite. These dependencies need to be understood and incorporated into models if one wants to predict the effectiveness of bentonite as a barrier to radionuclides in a nuclear repository. In this work, we present a benchmark problem for reactive transport simulators based on a flow-through experiment carried out on a saturated bentonite core. The measured effluent composition shows the complex interplay of species transport in a charged medium in combination with sorption and mineral precipitation/dissolution reactions. The codes compared in this study are PHREEQC, CrunchFlow, FLOTRAN, and MIN3P. The benchmark problem is divided into four component problems of increasing complexity, leading up to the main problem which addresses the effects of advective and diffusive transport of ions through bentonite with explicit treatment of electrostatic effects. All codes show excellent agreement between results provided that the activity model, Debye-Huckel parameters, and thermodynamic data used in the simulations are consistent. A comparison of results using species-specific diffusion and uniform species diffusion reveals that simulated species concentrations in the effluent differ by less than 8 %, and that these differences vanish as the system approaches steady state.