AbstractThe value of mechanistic ecosystem modelling has long been appreciated, and in connection with trait-based approaches it has recently stimulated a more process-based understanding of adaptive capacities and trade-offs. Notwithstanding recent advances, even sophisticated state-of-the-art models of plankton ecosystems, some of which include hundreds of idealized species, do not accurately represent the great biodiversity of plankton, or the associated flexible adaptive response of plankton communities. We build on previous reviews to suggest that it may be necessary to discard some common assumptions and try new approaches in order to construct models that can make new and testable predictions about the “adaptive capacity” of plankton ecosystems. Major challenges remain unresolved for modelling interacting communities of producers and consumers. Rather than the common approach of mixing and matching existing model components, each laden with its own legacy assumptions, we suggest that a judicious combination of innovative, mechanistic approaches that combine traits and trade-offs will likely better address such challenges.