In a world where floods are an important hazard, the incentive for flood forecasting, promoted by the advances in earth observations’ quality, is growing. In this context, while models are developed and used primarily on a national level, it has appeared essential to share data and models internationally. The present research inscribes itself into this scope. Seven years of hourly simulations and two weeks of hourly hindcasts with a lead time of five days, with and without an error correction procedure, were executed using the Delft-FEWS forecasting platform for four flood forecasting models and five sub-basins of the Meuse River basin. These models included two lumped conceptual models, the HBV-96 model, operationally used in the Netherlands and in Germany and the GR4H model, the hourly version of the GR4J model, used in France; and two fully-distributed physically based models, the wflow hbv and wflow sbm models developed at Deltares, the Netherlands. The quality of the simulations and of the hindcasts produced by each model for each of the five sub-basins was analysed and compared based on a combination of graphical techniques and performance metrics. This intercomparison framework brought to light several results. It has been shown that the Regnie interpolation method improved the HBV-96 model simulation performance for high flows compared to the Kriging interpolation method, highlighting the importance of data interpolation. Furthermore, the conceptual HBV-96 and GR4H models produced more reliable simulations and hindcasts than the physically-based wflow models. Finally, the use of high resolution input data did not show any clear improvement of the simulations and hindcasts quality. As part of a larger project, this research has set the starting point for further investigations regarding the disparities in the models results identified.