The objectives of this study were to assess and validate previously published prediction models with an independent dataset and to expose the power and limitations of linear regression models for predicting biomethane potential. Two datasets were used for the validation, one with all individual samples and one with the average values of each cultivar. The results revealed similar performances of all four models for the individual samples. The methane yields of the cultivars were predicted more accurately than the methane yields of the individual samples. The grassland specific model predicted the variation in the dataset with an R-2 of 0.84 and the slope of the regression line was equal to 1.0. Linear regression models are suitable to depict the variation in methane yield and for substrate ranking. However, the prediction error of the absolute values may be high since systematic external effects cannot be determined by a regression model.