Rainfall-runoff models are widely applied to simulate flow from rainfall. Lumped parameter models have the advantage of fast calculation with a small amount of input data requirement, but it cannot illustrate the variation of hydro-meteorology features by space, so it is only suitable for small basins. The distributed parameter models provide good description of spatial variation both in the basin responses and hydrological meteorological factors, but they require large amounts of data and longer computational time. This research builds a semi-distributed model combining the advantages of two groups of lumped and distributed parameter models based on the NAM model end the Muskingum routing. The model is applied in order to simulate daily runoff for the medium scale basins (Nong Son, Thanh My) and large scale basin (Cung Son). The Nash coefficient of calibration and verification processes are higher than 0.80, showing the effectiveness and the good applicability of the model.