Four gravel unconstrained flow experiments were modelled with the DAN-3D and RASH-3D codes. Both the codes, based on continuum mechanics, were developed for the propagation of rapid landslides, like rock avalanches. The codes were at first run to back-calculate the dynamic basal friction angle (frictional rheology) in order to model the runout of one of the experiments. The best fit value was then compared with the dynamic basal friction angle measured with a tilting test and finally applied for the modelling of the three other experiments. The back-calculated frictional parameter is different for the two codes and higher than the measured dynamic basal friction angle. This value can then be used to model the runout of other experiments involving a change of volume or falling height. On the other hand, in case of a modification of the slope angle, the dynamic basal friction angle has to be redefined. It seems thus difficult to use a single value of the basal friction angle to model experiments on various topographical profiles.