To analyse the broad-scale behaviour of 15 global models of the terrestrial biosphere, we evaluated the sensitivity of simulated net primary productivity (NPP) to spatial and seasonal variations in precipitation, temperature and solar radiation, and to the Normalized Difference Vegetation Index (NDVI). For annual NPP estimates, the models’ sensitivities to climate were the most similar in regions where NPP was not limited by precipitation. The largest differences in sensitivities occurred in regions where NPP was limited by both temperature and precipitation. Water use efficiencies within the models were relatively constant across latitudes so that higher correlations occurred between the latitudinal distribution of NPP and precipitation than with the other climate variables. The sensitivities of NPP estimates to solar radiation varied considerably with latitude. The largest differences in temperature sensitivity among NPP estimates occurred in the northern latitudes (50°N–70°N), i.e. the zone with the shortest active growing seasons. The sensitivity of NPP estimates to climate also varied seasonally. At the beginning and end of the active growing season in the boreal zone, monthly NPP estimates of all models were the most sensitive to temperature. In the tropics, sensitivities to climate varied widely among and within models. Seasonal changes in water balance and the structure of the vegetation canopy, as reflected by seasonal changes in NDVI, modified the sensitivity of NPP to climate in both boreal and tropical zones. Because these are both highly productive regions sensitive to climate change, continued investigations and better validation of models are necessary before we can fully understand and predict changes in ecosystem structure and function under various climatic conditions.