The fourth version of the Community Climate System Model (CCSM4) has been developed and released to the climate community. This paper describes the improvements and developments in all CCSM components, including the atmosphere, ocean, land, and sea ice components. The preindustrial control runs using both 1° and 2° resolutions in the atmosphere and land components are compared to the previous version, CCSM3, showing improvements in sea surface temperature biases and El Niño-Southern Oscillation variability. The ocean component has been enhanced with a new overflow parameterization, improving the Gulf Stream path and the North Atlantic Ocean meridional overturning circulation. The land component has been updated with a new hydrology scheme, improving the annual cycle of water storage. The sea ice component uses more realistic albedos, leading to an improved Arctic sea ice concentration. An ensemble of twentieth-century simulations shows a good match to observed September Arctic sea ice extent from 1979 to 2005. However, the globally averaged surface temperature increase between 1850 and 2005 in CCSM4 is larger than observed by about 0.4°C, possibly due to the absence of indirect aerosol effects. Despite these improvements, significant biases remain, such as precipitation distribution in the tropical Pacific Ocean and low cloud amount in the Arctic.The fourth version of the Community Climate System Model (CCSM4) has been developed and released to the climate community. This paper describes the improvements and developments in all CCSM components, including the atmosphere, ocean, land, and sea ice components. The preindustrial control runs using both 1° and 2° resolutions in the atmosphere and land components are compared to the previous version, CCSM3, showing improvements in sea surface temperature biases and El Niño-Southern Oscillation variability. The ocean component has been enhanced with a new overflow parameterization, improving the Gulf Stream path and the North Atlantic Ocean meridional overturning circulation. The land component has been updated with a new hydrology scheme, improving the annual cycle of water storage. The sea ice component uses more realistic albedos, leading to an improved Arctic sea ice concentration. An ensemble of twentieth-century simulations shows a good match to observed September Arctic sea ice extent from 1979 to 2005. However, the globally averaged surface temperature increase between 1850 and 2005 in CCSM4 is larger than observed by about 0.4°C, possibly due to the absence of indirect aerosol effects. Despite these improvements, significant biases remain, such as precipitation distribution in the tropical Pacific Ocean and low cloud amount in the Arctic.