We have built the first comprehensive global three-dimensional model of δ18O in atmospheric CO2. The constructed model goes beyond all other approaches made until now, by simulating the diurnal variations and transport of CO2, δ18O of water, and δ18O of CO2. The CO18O fluxes are thereby dependent on the atmospheric CO18O composition. We have validated the model surface processes, showing that it compares well to other estimates and measurements of NPP, NEE, and stomata-internal CO2 mixing ratio (ci), except for high northern latitudes. Here, the model is considerably lower in NPP and higher in ci than other model estimates. However, estimates derived indirectly from observations tend to support our model findings. The water isotopes of rain are reproduced very well at all latitudes. The soil bucket model used in the model integrates incoming rain in one single value. The bucket approach overattenuates the isotopic variations of rain, and hence our isotopic source signature of respiration shows almost no seasonal cycle and is thus isotopically too depleted during summer.