Our understanding of climate change impacts on boreal forest net primary productivity (NPP) largely rely on terrestrial biosphere models (TBMs). TBMs characterize forest NPP through predictions of plant carbon fluxes and allocation from climatic data, physiological parameters and process logic. TBMs’ outcomes may be highly variable, a major source of uncertainty being the formulation of ecophysiological processes and their parameterization. In our study, we compared patterns of growth response to climate across Canada’s boreal forest obtained either from a spatially-extensive tree-ring network to or from NPP outputs of fourteen (TBMs). We found that tree growth varied along a southwest/northeast longitudinal gradient of decreasing water limitation to increasing energy limitation. All but two TBMs partially reproduced this spatial gradient of growth limitations. Amongst the climatic variables tested, there was a strong mismatch in the strength and direction of growth responses to summer temperature, largely positive for NPP but negative for tree growth. The negative relationship impact of summer temperature on growth mainly took place during the year prior to ring formation; no such lagged effect was found in the NPP responses. Much of the temperature signal in tree growth could be explained by seasonal fluctuations in atmospheric vapor pressure deficit and soil moisture availability, suggesting a dominant indirect effect of temperature on boreal tree growth via these two variables. Our results suggest that the formulation of processes representing growth responses to climate and their parameterization are not adequately representing productivity of Canada’s boreal forest in a majority of tested TBMs.