The Community Climate System Model Version 3 (CCSM3) is a coupled climate model designed to simulate past, present, and future climates with high realism and stability over a wide range of spatial resolutions. CCSM3 incorporates significant improvements in physical parameterizations, including cloud processes, aerosol radiative forcing, land-atmosphere fluxes, ocean mixed layer processes, and sea ice dynamics. These enhancements have reduced or eliminated systematic biases in the mean climate produced by previous versions of the model. Key improvements include better simulations of sea ice thickness, polar radiation budgets, tropical sea surface temperatures, and cloud radiative effects. However, systematic biases remain in ocean-atmosphere fluxes in coastal regions, ENSO variability, tropical ocean precipitation, and continental precipitation and surface air temperatures. The model's stability and long-term behavior are also discussed, showing that it can produce stable climate simulations over several millennia without ad hoc adjustments. Future work aims to further refine the model to address remaining scientific challenges.The Community Climate System Model Version 3 (CCSM3) is a coupled climate model designed to simulate past, present, and future climates with high realism and stability over a wide range of spatial resolutions. CCSM3 incorporates significant improvements in physical parameterizations, including cloud processes, aerosol radiative forcing, land-atmosphere fluxes, ocean mixed layer processes, and sea ice dynamics. These enhancements have reduced or eliminated systematic biases in the mean climate produced by previous versions of the model. Key improvements include better simulations of sea ice thickness, polar radiation budgets, tropical sea surface temperatures, and cloud radiative effects. However, systematic biases remain in ocean-atmosphere fluxes in coastal regions, ENSO variability, tropical ocean precipitation, and continental precipitation and surface air temperatures. The model's stability and long-term behavior are also discussed, showing that it can produce stable climate simulations over several millennia without ad hoc adjustments. Future work aims to further refine the model to address remaining scientific challenges.