Radiocarbon is a powerful tracer to determine the state of the climate system as well as to investigate past environmental changes. However, 14C observations become increasingly scarce the further we look backward in time. General circulation models (GCMs) may help to interpret scattered 14C records by dynamically filling in spatio-temporal gaps. Here, we give two examples for such a GCM-based approach in which we apply an ocean GCM to study issues related to the last deglaciation. First we show results of equilibrium simulations indicating that during the Last Glacial Maximum, the overturning circulation of the ocean and the 14C partitioning between the various marine reservoirs were quite different from today. Then we will focus on transient simulations in which we assess the coevolution of marine 14C records and atmospheric 14C reconstructions between 21 and 14 cal ka BP. These simulations employ an iteration scheme in which atmospheric 14C reconstructions are reanalyzed and readjusted in a self-consistent way. An example considering data from the Caribbean Sea leads to enhanced variability of atmospheric Δ14C by +/- 30 ‰, and to an increase of the mysterious drop of atmospheric Δ14C between 17.5 and 14.5 cal ka BP by about 20‰. A by-product of this ‘reanalysis’ are crude estimates of cosmogenic 14C production fluxes which, for the given example, are in the lower range of reconstructions.