Following a three-dimensional inverse modeling study on atmospheric CO mixing ratios <Bergamaschi et al., this issue>, we present the expansion of our model for the treatment of the stable isotope ratios 13C/12C and 18O/16O in carbon monoxide. The individual isotopomers were included in the model as independent tracers, and the inversion scheme was extended, allowing the simultaneous optimization (with respect to observational data) of modeled atmospheric mixing and isotope ratios. Observational data of 13C/12C and 18O/16O ratios were taken from a set of five globally distributed sites, all of which exhibit pronounced seasonal cycles and which as a whole clearly define large latitudinal gradients. Incorporation of 13C/12C ratios in the inversion resulted in a clear constraining of the total amount of CO arising from CH4 oxidation. The present study suggests an average CO yield of 86% (80--88%), provided that (1) the reaction of CH4+Cl has no significant impact on the Δ13C of the resulting CO and (2) CO from the ocean helps to balance the Δ13C in the Southern Hemisphere. Otherwise, a further reduced CO yield would be necessary, as low as 71% in case of an average kinetic isotope effect of 1.013 <Lowe et al., 1999> or 69% in case of a Δ13C value of -25.0? for the oceanic source. Despite a lack of experimental investigations on 18O/16O in CO arising from CH4 or nonmethanehydrocarbons (NMHC) oxidation, inclusion of 18O/16O into the inversion gives further significant constraints on the global CO cycle. It is shown that apart from the large technological CO source, additional sources with dominant emissions in the Northern Hemisphere are required, such as biogenic emissions (either direct or via oxidation of biogenic NMHCs) or anthropogenic NMHCs, in order to reproduce observed atmospheric 18O/16O ratios. The extended inversion scheme allows CO budgets to be derived, which reproduce, within two standard deviations of observational data, simultaneously CO mixing ratios (from the National Oceanic and Atmospheric Administration Climate Monitoring and Diagnostics Laboratory network), 13C/12C, and 18O/16O ratios. Incorporation of stable isotopes renders inversion results much more robust compared to inversions of CO mixing ratios only. ¿ 2000 American Geophysical Union