The objective of this study was to predict the co-grinding of calcined clay and clinker in an industrial circuit. The N perfect mixers in series approach was used to model a ball mill. The input parameters (grindability of calcined clay and clinker) were first determined using a laboratory batch ball mill. The model was then successfully validated at the industrial level for clinker. The grindability of each of the calcined clay minerals (mainly quartz, metakaolin and muscovite) was estimated as follows: the calcined clay, ground in the laboratory mill at different grinding times, was classified at different finenesses, and the resulting composition was measured for each size fraction. The individual Particle Size Distributions (PSD) of each clay phase was then determined using a deconvolution algorithm, and their grindability parameters were inferred accordingly. The grinding model can be used for example to predict the impact of industrial parameters (feed rate, separator speed…) on the PSD of the ground product and on the specific energy. Moreover, the model gives access to the PSD of all the individual clay minerals. This helps predict, e.g. the reactivity of the calcined clay in the interground cement, or estimate the Size-Weighted Respirable Fraction (SWeRF) of crystalline silica in the cement.