Chemical weed control before crop and weed emergence is a systematic practice in winter oilseed rape crops in France. It would be profitable both for farmers and the environment to predict the level of weed infestation early on in the growing season and to control weeds only when necessary using post-emergence weed control. The objective of this paper was to develop and evaluate simple models to predict weed biomass in oilseed rape crops. The model input variables were related to weed population characteristics and farmers' practices. The models can be used to classify oilseed rape plots into two categories: plots with a level of weed infestation above a threshold or those with level of weed infestation below a threshold. A data set including 3 years of experiments, conducted across several regions in France, was used to estimate the parameters and to evaluate the models. High values of sensitivity and specificity were obtained when weed biomass was predicted as a function of sowing date, type of soil tillage, soil mineral nitrogen, crop density, weed density at emergence, and main characteristics of the most abundant weed species. Model performance strongly decreased when input variables related to the weed population were not taken into account. The best models correctly classified 90% of the plots with high weed infestation and 64% of the plots with low weed infestation.