Several pelagic ecosystem models now allow for flexibility in the elemental composition of the modelled plankton functional types. Often they are based on some variant of the model for phytoplankton photoacclimation by Geider et al. (1998) that has been shown to reproduce C:N and C:Chl ratios in laboratory experiments both for constant nutrient and light conditions, and for a transition to nutrient limitation well. Based on that success, the predicted stoichiometries in ecosystem models employing the Geider model or one of its competitors have not received much attention. Here we first investigate how well the two physiologically-based models by Geider et al. (1998), and by Pahlow (2005) are able to describe observed changes in biomass stoichiometry in a laboratory experiment with a more dynamic nutrient regime, namely a sudden resupply of nitrogen after a bloom, as would be expected after a brief mixing event. We show that both the Geider et al. (1998), and the Pahlow (2005) model are able to describe shifts on C:N stoichiometry over the course of a bloom, albeit not equally well, but that both have problems describing the recovery from nutrient starvation. We then discuss the patterns of phytoplankton stoichiometry and their sensitivity to some assumptions on physiology in a global ecosystem model. Our model allows for flexible stoichiometry in both non-diatoms and diatoms, with an extension of the Geider model for diatom Si:C variations. Globally, largest deviations of the C:N ratio from Redfield are modelled in the nutrient-depleted subtropical gyres, but they contribute little to vertical export. Lab-observed changes in C:Chl ratio are described well by the Geider model. Modelled global patterns in C:Chl are consistent with a number of field- and satellite-based observations. It remains a challenge, however, to distinguish between variations in stoichiometry due to physiological acclimation and to community composition.