Process-based forest growth models with spatially explicit representation are relevant tools to investigate innovative silviculture practices and/or climate change effects because they are based on key ecophysiological processes and account for the effects of local competition for resources on tree growth. Such models are rare and are often calibrated for a very limited number of species and rarely for mixed and/or uneven-aged stands, and none are suitable for the temperate forests of Québec. The aim of this study was to calibrate and evaluate HETEROFOR (HETEROgeneous FORest), a process-based and spatially explicit model based on resource sharing, for 23 functionally diverse tree species in forest stands with contrasting species compositions and environmental conditions in southern Quebec. Using data from the forest inventory of Quebec, we evaluated the ability of HETEROFOR to predict the short-term growth (5–16 years) of these species at the tree and stand levels and the long-term dynamics (120 years) of red and sugar maple stands. The comparison between the prediction quality of the calibration and evaluation datasets showed the robustness of the model performance in predicting individual-tree growth. The model reproduced correctly the individual basal area increment (BAI) of the validation dataset, with a mean Pearson's correlation coefficient of 0.56 and a mean bias of 18 %. Our results also highlighted that considering tree position is of importance for predicting individual-tree growth most accurately in complex stands with both vertically and horizontally heterogeneous structures. The model also showed a good ability to reproduce BAI at the stand level, both for monospecific (bias of −3.7 %; Pearson's r=0.55) and multi-species stands (bias of −9.1 %; Pearson's r=0.62). Long-term simulations of red maple and sugar maple showed that HETEROFOR was able to accurately predict the growth (basal area and height) and mortality processes from the seedling stage to the mature stand. Our results suggest that HETEROFOR is a reliable option to simulate forest growth in southern Quebec and to test new forestry practices under future climate scenarios.