A model predicting the dynamics of herbage biomass, structure and digestibility in managed permanent grasslands (Jouven et al., 2006a) was tested on data obtained over three successive vegetation cycles and two years from three contrasting grasslands: based on the classification given in Cruz et al. (2002), plot one (86% grasses) was dominated by functional group A species (Lolium perenne), plot two (68% grasses) was dominated by group B (Dactylis glomerata) and group b species (Holcus mollis), and plot three (only 39% grasses) was dominated by group b (Agrostis capillaris) and group C species (Festuca rubra). For the three types of grasslands, the precision of the simulation of biomass production was good for the first vegetation cycle but lower for the second and the third cycles. Digestibility was precisely simulated during the first cycle for grasslands one and two but was overestimated for plot three. It is concluded that model could be improved by better taking into account the seasonal modulation of growth, and by better knowledge of the herbage quality of species from functional groups b and C.