The paper discusses the development and operational implementation of the second version of the NCEP Climate Forecast System (CFSv2), which was launched in March 2011. CFSv2 represents significant improvements over its predecessor, CFSv1, in all aspects of data assimilation and forecast model components. A 32-year coupled reanalysis (1979–2010) was conducted to provide initial conditions for a comprehensive reforecast period from 1982 to 2010, aimed at calibrating and assessing the skill of subseasonal and seasonal predictions. The operational implementation of CFSv2 ensures continuity in climate records and provides valuable retrospective forecasts for various applications, including water management, agriculture, transportation, energy use, and hurricane season prediction. Key improvements in CFSv2 include enhanced atmospheric and land surface models, an upgraded soil model, an interactive sea ice model, and historically prescribed CO2 concentrations. These enhancements have led to significant improvements in subseasonal and seasonal forecasts, such as increased skill in Madden-Julian Oscillation (MJO) forecasts, better 2-meter temperature predictions over the United States, and improved global sea surface temperature forecasts. The paper also presents a detailed evaluation of CFSv2's performance, comparing it with CFSv1 and other models used in the U.S. National Multimodel Ensemble (NMME). CFSv2 shows improved skill in subseasonal forecasts, particularly in MJO predictions, and in seasonal forecasts of 2-meter temperature over the Northern Hemisphere. Probabilistic seasonal predictions, including tercile forecasts of Niño-3.4 sea surface temperature, also demonstrate enhanced reliability and resolution. Additionally, the paper discusses diagnostics of model behavior, including systematic errors, surface water budgets, and sea ice predictions. While sea ice predictions show some biases, the model generally captures the seasonal cycle and interannual variability. The paper concludes with a discussion of long-term integrations, showing that CFSv2 produces consistent warming trends and sufficient spread to cover observed climate changes.The paper discusses the development and operational implementation of the second version of the NCEP Climate Forecast System (CFSv2), which was launched in March 2011. CFSv2 represents significant improvements over its predecessor, CFSv1, in all aspects of data assimilation and forecast model components. A 32-year coupled reanalysis (1979–2010) was conducted to provide initial conditions for a comprehensive reforecast period from 1982 to 2010, aimed at calibrating and assessing the skill of subseasonal and seasonal predictions. The operational implementation of CFSv2 ensures continuity in climate records and provides valuable retrospective forecasts for various applications, including water management, agriculture, transportation, energy use, and hurricane season prediction. Key improvements in CFSv2 include enhanced atmospheric and land surface models, an upgraded soil model, an interactive sea ice model, and historically prescribed CO2 concentrations. These enhancements have led to significant improvements in subseasonal and seasonal forecasts, such as increased skill in Madden-Julian Oscillation (MJO) forecasts, better 2-meter temperature predictions over the United States, and improved global sea surface temperature forecasts. The paper also presents a detailed evaluation of CFSv2's performance, comparing it with CFSv1 and other models used in the U.S. National Multimodel Ensemble (NMME). CFSv2 shows improved skill in subseasonal forecasts, particularly in MJO predictions, and in seasonal forecasts of 2-meter temperature over the Northern Hemisphere. Probabilistic seasonal predictions, including tercile forecasts of Niño-3.4 sea surface temperature, also demonstrate enhanced reliability and resolution. Additionally, the paper discusses diagnostics of model behavior, including systematic errors, surface water budgets, and sea ice predictions. While sea ice predictions show some biases, the model generally captures the seasonal cycle and interannual variability. The paper concludes with a discussion of long-term integrations, showing that CFSv2 produces consistent warming trends and sufficient spread to cover observed climate changes.