Many-core optimization of the Community Earth System Model The Application Scalability and Performance group at the National Center for Atmospheric Research has made significant progress in reducing the execution time of multiple pieces of the Community Earth System Model (CESM) on both Xeon and Xeon Phi architectures. Transformations to improve vectorization and cache locality typically reduces execution time for pieces of CESM on both Xeon and Xeon Phi by a factor of two. Progress has been made possible through the use of an automated kernel extraction tool and extensive collaboration with multiple institutions and several computer vendors. An overview of our approach and progress is provided.