Nuclear waste repositories will use a significant quantity of cement: for the construction of access drifts, disposal cells and concrete plugs, and as containment material for low- to intermediate-level waste. Several European countries have chosen claystone formations as possible host rocks (Landais, 2006). Numerous cement/clay interfaces will thus be present in a radioactive-waste repository. Due to contrasting geochemical conditions (including Eh, pH, solution composition), these interfaces are subjected to steep concentration gradients and are highly reactive. Predicting long-term changes (1,000 to 100,000 years) in these cementitious and clayey materials is thus crucial for assessing the behaviour of such infrastructures. Experiments cannot provide sufficiently reliable information over such long time scale. Although natural and archaeological analogues can be very helpful, modelling is the unique tool to analyse and test different evolution scenarios. In order to build a better confidence in such calculations, it is of paramount importance to demonstrate that the results obtained are not dependent on the choice of the numerical reactive transport code to perform the simulation. In order to address this issue, a benchmark problem, divided into three steps with increasing geochemical refinement, has been set up (Marty et al., submitted). In all cases, the solutes transport across the interface between clayey host rock and concrete is diffusion driven and a 1D radial geometry and isothermal conditions (25°C) have been both considered. Both materials are full saturated. The first step of the benchmark only considers porewater solutions and the clayed host rock which is only modelled by an exchanger. The second step introduces the full mineralogy for both the concrete and the claystone considering slow kinetics rates for mineral dissolution-precipitation reactions, whereas the third one focuses on fast reaction rates. Seven international teams have been involved in this benchmarking exercise. All reactive transport codes used (TOUGHREACT, PHREEQC with two different ways of handling transport, CRUNCHFLOW, HYTEC, ORCHESTRA, MIN3P-THCm) gave very similar patterns in terms of predicted solute concentrations and of minerals distribution evolution (Fig. 1). The benchmarking exercise demonstrates that reactive transport tools have reached such a level of maturity as to be confidently used in support of long-term performance assessments.