Improving the spatial resolution and forecasting accuracy of rainfall-runoff model is one of the hotspots and difficulties in flood forecasting and hydrological research. To address this problem, a high-resolution rainfall-runoff model, Grid-XAJ-SATIN, considering the spatial combination of infiltration-excess and saturation-excess runoff is proposed in this paper, which is based on the existing study of the grid-based rainfall-runoff models. Several aspects of this model are discussed including its principle and structure, processing of the geographical information, spatial distribution of parameters, model forcing data and assimilation of basin states. The model is then applied to both humid and semi-humid catchments for flood simulation. The simulation results show that the accuracy of the flood forecasting is well realized in both conditions. Moreover, this model is demonstrated to be able to forecast not only the hydrograph at the catchment outlet but also the streamflow at any interior grid cells.