Prediction of plastic properties of blends is a prerequisite for optimal blend design since coke quality indices depend on them. 108 blends and their 121 components coals were characterized by Gieseler plastometry and used to test different models of prediction of the logarithm of maximum fluidity (LMF) of blends. Best results are obtained by the method of reconstruction of blend plastograms consisting in weight averaging the logarithm of fluidity of the component coals whose plastometric curves are approximated by two half-parabolas. Its performances improve with increasing number of components coals, probably due to the decrease in the resultant of all the possible interactions between coals. The softening and resolidification temperatures of blends can be expressed as a function of an "interaction parameter" which quantifies the fusion/solidification-retardant/accelerant effect induced by interactions between blend components. Because this interaction parameter does not change too much between softening and resolidification, plastic range (PR) is less sensitive to interactions than LMF and can be satisfactorily modelled by additivity but, from the only point of view of performance, LMF prediction is better than the PR one. (C) 2017 Elsevier Ltd. All rights reserved.