An automatic in vitro gas production technique was evaluated for predicting in vivo fiber (NDF) digestibility and effective first-order digestion rate of potentially digestible NDF (pdNDF) of 15 grass silages. Observed in vivo NDF digestibility of the silages harvested at different stages of maturity during 3 yr was determined by the total fecal collection in sheep fed at the maintenance level of intake. Isolated grass silage NDF was incubated for 72 h in the presence of rumen fluid and buffer to determine the pdNDF digestion kinetics based on cumulative gas production profiles. The digestion kinetic parameters were estimated by a 2-pool Gompertz function. The estimated parameter values were then used in a 2-compartment mechanistic rumen model to predict the in vivo digestibility of pdNDF. A total compartmental mean residence time of 50 h was used in the model, and a further assumption of the distribution of the residence time between the rumen nonescapable and escapable pools in a ratio of 0.4:0.6 was made. To make a distinction between potentially digestible and indigestible NDF, the potential extent of NDF digestion was determined by a 12-d ruminal in situ incubation. The model-predicted in vivo NDF digestibility accurately and precisely (root mean square error = 0.013 units, R(2) = 0.99). Effective first-order digestion rate was estimated from the predicted pdNDF digestibility, and the values were compared with those calculated from the in vivo pdNDF digestibility using the same passage kinetic parameters. The predicted effective first-order digestion rate was strongly correlated with digestion rate estimates derived from in vivo data (root mean square error = 0.006/h, R(2) = 0.86). It can be concluded that a simple first-order digestion rate can be estimated from a complicated gas production kinetic model including 6 parameters. This rate constant can be used in continuous steady-state dynamic mechanistic rumen models predicting the nutrient supply to the host animal.