A state-of-the-art complex marine ecosystem model, PlankTOM5.2, simulating the distribution of five plankton functional types (PFTs; mixed phytoplankton, diatoms, coccolithophores, micro and mesozooplankton), was implemented separately in two medium resolution (order 1°) global ocean general circulation models (OGCMs), NEMO and OCCAM. In each case, identical formulations and parameter values were used in the ecosystem model, as well as the same biogeochemical forcing (photosynthetically active radiation and input of nutrients by atmospheric dust at the ocean surface) and initial conditions. The two physical models were forced with essentially the same surface boundary conditions, subject to interpolation between the two model grids, and simulations were undertaken for years 1990-1994. Sensitivity of the ecosystem model and associated biogeochemical fields was assessed with respect to differences in the ocean physics of the two OGCMs. Globally integrated bulk properties, notably annual mean primary production and total phytoplankton biomass, were similar in each case (although export showed regional differences between the models), as well as being generally consistent with available observations. In contrast, predicted distributions of individual PFTs varied markedly between the two simulations. Diatoms and microzooplankton were for example predicted to dominate in the North Atlantic, North Pacific and Southern Oceans in OCCAM because of relatively strong mixing which supplied increased nutrients that favoured diatom production and which also increased the mortality of their main predators, the mesozooplankton, which struggled to recover their numbers over winter as phytoplankton were depleted. In NEMO, lower mixing led to a community structure in which mixed phytoplankton and mesozooplankton were the main PFTs in the same areas. Regions of dominance by coccolithophores and mixed phytoplankton were predicted in the tropics, the former group being more extensive in the subtropics in NEMO, with distributions depending primarily on the size and extent of upwelling and downwelling regions predicted by the OGCMs. Results highlight the need for accuracy both when formulating the equations for, and parameterising, PFTs in models, and moreover in the representation of the physico-chemical environment.