Global?scale models of atmospheric chemistry (GACMs) “mix” biomass burning emissions into grid boxes with horizontal scales of 10–200 km. This ignores the complex nonlinear transformations that take place in the young smoke plumes. Here we use a new gas? and aerosol?phase chemistry model called Aerosol Simulation Program (ASP) and a 3?D Eulerian smoke plume model to simulate the fluid dynamics, radiative transfer, gas?phase chemistry, and aerosol?phase chemistry of the Timbavati smoke plume observed during SAFARI 2000. We then compare the results of the 3?D plume model with those of an Eulerian box model, which is used as an analog for the large grid boxes of GACMs. The 3?D plume model matched the observed plume injection height but required a large minimum horizontal diffusion coefficient to match the observed horizontal dispersion of the plume. Absorption and scattering by smoke aerosols reduced the modeled photolysis rates in the plume by 10–20%. Increasing the heterogeneous production of HONO and H2SO4 in the model and including uncharacterized organic species using monoterpenes as a proxy compound improves the model?observation match. Direct measurements of OH in the smoke plumes would be an excellent way to determine if heterogeneous production of HONO is taking place. The automatic dilution of smoke plume emissions into the large grid boxes of global models can result in large errors in predicted concentrations of O3, NOx and aerosol species downwind. We discuss several potential approaches that could reduce these errors.