The automatization of open-die forging technology is essential for cheaper production and better quality products. The current models for the prediction of spreading are too general and, consequently, not sufficiently accurate, which was confirmed by using Hot Forging Simulator software (HFS). The metods applied for the model development that include more influential parameters, are the following: the registration of parameters and measurements of workpiece spreading during the laboratory physical simulations; industrial measurements; the analysis of local material flow with the FE method; the treatment of measurement results and analysis with neuronal networks. Two models for the prediction of free spreading were proposed: prediction in the form of a function and prediction by applying neuronal networks. The accuracy of the predictions with both new models is much better than that with the current model.