Ocean biogeochemical models are key tools for both scientific and operational applications. Nevertheless the cost of these models is often expensive because of the large number of biogeochemical tracers. This has motivated the development of multi-grid approaches where ocean dynamics and tracer transport are computed on grids of different spatial resolution. However, existing multi-grid approaches to tracer transport in ocean modelling do not allow the computation of ocean dynamics and tracer transport simultaneously. This paper describes a new multi-grid approach developed for accelerating the computation of passive tracer transport in the Nucleus for European Modelling of the Ocean (NEMO) ocean circulation model. In practice, passive tracer transport is computed at runtime on a grid with coarser spatial resolution than the hydrodynamics, which reduces the CPU cost of computing the evolution of tracers. We describe the multi-grid algorithm, its practical implementation in the NEMO ocean model, and discuss its performance on the basis of a series of sensitivity experiments with global ocean model configurations. Our experiments confirm that the spatial resolution of hydrodynamical fields can be coarsened by a factor of 3 in both horizontal directions without significantly affecting the resolved passive tracer fields. Overall, the proposed algorithm yields a reduction by a factor of 7 of the overhead associated with running a full biogeochemical model like PISCES (with 24 passive tracers). Propositions for further reducing this cost without affecting the resolved solution are discussed.