This paper presents a new open-source "GBRV" ultra-soft pseudopotential library optimized for high-throughput density-functional theory (DFT) calculations. The GBRV library includes accurate ultra-soft pseudopotentials generated using the Vanderbilt pseudopotential generation code. The authors benchmark the GBRV potentials, as well as two other pseudopotential sets (VASP PAW and PSLIB), against all-electron calculations to validate their accuracy. The results show that the GBRV potentials perform well in metallic, ionic, and covalent bonding environments, with most lattice constants within 0.2% of all-electron results and most bulk moduli within 5%. The GBRV potentials are more robust than the VASP and PSLIB sets, especially for structures with multiple elements in covalent and ionic environments. The GBRV potentials are also more accurate for magnetic moments than the VASP set. The GBRV library is available at http://physics.rutgers.edu/gbrv and can be used directly with QUANTUM ESPRESSO and ABINIT. The authors also discuss the design criteria for the GBRV library, including the use of a low plane-wave cutoff and the inclusion of semi-core states for transferability. The GBRV potentials are designed to run at a plane-wave cutoff of 40 Ry and a charge-density cutoff of 200 Ry, which are significantly lower than many other pseudopotential sets. The GBRV library is open-source, allowing for easy replication and modification of the potentials. The authors conclude that the GBRV potentials provide better accuracy and robustness than other pseudopotential sets for high-throughput DFT calculations.This paper presents a new open-source "GBRV" ultra-soft pseudopotential library optimized for high-throughput density-functional theory (DFT) calculations. The GBRV library includes accurate ultra-soft pseudopotentials generated using the Vanderbilt pseudopotential generation code. The authors benchmark the GBRV potentials, as well as two other pseudopotential sets (VASP PAW and PSLIB), against all-electron calculations to validate their accuracy. The results show that the GBRV potentials perform well in metallic, ionic, and covalent bonding environments, with most lattice constants within 0.2% of all-electron results and most bulk moduli within 5%. The GBRV potentials are more robust than the VASP and PSLIB sets, especially for structures with multiple elements in covalent and ionic environments. The GBRV potentials are also more accurate for magnetic moments than the VASP set. The GBRV library is available at http://physics.rutgers.edu/gbrv and can be used directly with QUANTUM ESPRESSO and ABINIT. The authors also discuss the design criteria for the GBRV library, including the use of a low plane-wave cutoff and the inclusion of semi-core states for transferability. The GBRV potentials are designed to run at a plane-wave cutoff of 40 Ry and a charge-density cutoff of 200 Ry, which are significantly lower than many other pseudopotential sets. The GBRV library is open-source, allowing for easy replication and modification of the potentials. The authors conclude that the GBRV potentials provide better accuracy and robustness than other pseudopotential sets for high-throughput DFT calculations.