We have compared vertical transport of temperature, anthropogenic CO2, natural radiocarbon (14C), and bomb 14C in a global box-diffusion model (B-D) and a three-dimensional (3-D) ocean general circulation model from the Geophysical Fluid Dynamics Laboratory. Our main objectives were (1) to test the eddy diffusion parameterization of large-scale vertical transport in ocean box models and (2) to assess the utility of bomb-produced and natural 14C observations to validate ocean models used to estimate anthropogenic CO2 uptake. From the 3-D model's distributions and fluxes of natural 14C, bomb 14C, and anthropogenic CO2, we have calculated apparent diffusivities (Kap) vertically over the global ocean that range mostly between 4000 and 8000 m2 yr-1. These Kap agree quantitatively with diffusivities found by fitting B-D models to observed distributions of natural and bomb 14C. We then used these sets of Kap in different runs of a global B-D model. Results from all B-D models runs matched to within 13% those from the 3-D model for global uptake of anthropogenic CO2 and bomb-14C penetration depth. Although Kap from 3-D simulations for bomb 14C vary with time, those from 3-D runs for anthropogenic CO2 are essentially constant. Still, we found nearly the same results with the B-D model when Kap from 3-D bomb 14C simulations are approximated as time invariant. The best agreement (within 3%) between 3-D CO2 simulations and B-D model runs was found when applying Kap derived from bomb 14C in the surface and from natural 14C in the deep. Agreement was worse when using Kap from 3-D simulations for anthropogenic CO2 itself, mostly because in this case deeper Kap could only be extrapolated from higher surface values. We have found it appropriate to study global oceanic uptake of anthropogenic CO2 with B-D model and to validate anthropogenic carbon uptake models using natural and bomb 14C observations. For bomb 14C in the 3-D model, convective transport was most important during 1955--1964 while atmospheric levels were rising; afterward, atmospheric levels drop, and advective overturning dominates as for natural 14C. Thus 14C seems less than ideal to validate the convective scheme of general circulation models.¿ 1997 American Geophysical Union