Plankton models form the core of marine ecosystem simulators, with uses from regional resource and ecosystem management to climate change projections. In this Perspective, we suggest that stronger alignment of models with empirical knowledge about plankton physiology, diversity and trophic roles will improve model utility and the reliability of their outputs regarding biodiversity, ecophysiology, trophic dynamics and biogeochemistry. We recommend key steps to resolve the disconnect between empirical research and simulation models accounting for well-established plankton processes with an aim to increase the utility of such models for applied uses. A central challenge is characterizing the complexity of plankton diversity and activity in ways that are amenable to model incorporation. We argue that experts in empirical science are best placed to advise the development of next-generation models to address these challenges, and we propose a series of actions to achieve that engagement, including involvement of these experts in the design and exploitation of plankton digital twins.