Abstract. Global chemistry-climate models (CCMs) play an important role in assessing the climate and air pollution implications of aerosols and chemically reactive gases. Evaluating these models under past conditions and constraining historical sources and sinks necessitates reliable records of atmospheric mixing ratios spanning preindustrial times. Such precious records were recently obtained for carbon monoxide (CO) documenting for the first time the evolution of this reactive compound over the industrial era. In this study, we compare the simulated atmospheric surface CO mixing ratios ([CO]) from two different sets of CCMs and emissions in the frame of CMIP5 and of CMIP6 (Coupled Model Intercomparison Project Phases 5 and 6) with recent bipolar ice archive reconstructions for the period spanning 1850 to present. We analyze how historical (1850–2014) [CO] outputs from 16 (Atmospheric Chemistry and Climate Model Intercomparison Project) models and 6 AerChemMIP (Aerosol Chemistry Model Intercomparison Project) models over Greenland and Antarctica are able to capture both absolute values and trends recorded in multi-site ice archives. While most models underestimate [CO] at high northern latitudes, a reduction in this bias is observed from ACCMIP to AerChemMIP exercises. Over the 1980–2010 CE period, trends in ice archive and firn air observations and AerChemMIP outputs align remarkably well at high northern and southern latitudes, indicating improved quantification of CO anthropogenic emissions and the main CO sink (OH oxidation) compared to ACCMIP. From 1850 to 1980 CE, AerChemMIP models and observations consistently show increasing [CO] in both the Northern Hemisphere (NH) and Southern Hemisphere (SH), suggesting a robust understanding of the CO budget evolution. However, a divergence in the [CO] growth rate emerges in NH between models and observations over the 1920–1975 CE period, attributed to uncertainties in CO emission factors (EF), particularly EF for RCO (Residential, Commercial and Others) and transportation sectors, although we cannot totally rule out that the CO record based on Greenland ice archives may be biased high by CO chemical production processes occurring in the ice prior the measurements (i.e., in situ CO production). In the Southern Hemisphere, AerChemMIP models simulate an increase in atmospheric [CO] from 1850 to 1980 CE closely reproducing the observations (22±10 ppb and 13±7 ppb, respectively). Such agreement supports CMIP6 biomass burning CO emission inventories which do not reveal a peak in CO emissions in the late 19th century. Furthermore, both SH models and observations reveal an accelerated growth rate in [CO] during 1945–1980 CE relative to 1980–1945 CE, likely linked to increased anthropogenic transportation emissions.