The CUBA library provides four new implementations of multidimensional numerical integration algorithms: Vegas, Suave, Divonne, and Cuhre. These algorithms are implemented in C, C++, Fortran, and Mathematica, with similar interfaces and easy interchangeability. All can integrate vector functions and have different approaches to numerical integration. Vegas uses importance sampling and is simple but not always efficient. Suave combines Vegas' importance sampling with subregion sampling, improving adaptability. Divonne uses stratified sampling and optimization techniques to find peaks efficiently. Cuhre is a deterministic algorithm using cubature rules for high precision. The algorithms have been tested and optimized for performance, with improvements in convergence and memory usage. The library is available for download and compilation, with detailed documentation for use in Fortran, C/C++, and Mathematica. The library is designed for high-dimensional integrals and is particularly effective for integrands with sharp peaks or oscillatory behavior. The algorithms are flexible, allowing users to specify integration parameters and options for different integration scenarios.The CUBA library provides four new implementations of multidimensional numerical integration algorithms: Vegas, Suave, Divonne, and Cuhre. These algorithms are implemented in C, C++, Fortran, and Mathematica, with similar interfaces and easy interchangeability. All can integrate vector functions and have different approaches to numerical integration. Vegas uses importance sampling and is simple but not always efficient. Suave combines Vegas' importance sampling with subregion sampling, improving adaptability. Divonne uses stratified sampling and optimization techniques to find peaks efficiently. Cuhre is a deterministic algorithm using cubature rules for high precision. The algorithms have been tested and optimized for performance, with improvements in convergence and memory usage. The library is available for download and compilation, with detailed documentation for use in Fortran, C/C++, and Mathematica. The library is designed for high-dimensional integrals and is particularly effective for integrands with sharp peaks or oscillatory behavior. The algorithms are flexible, allowing users to specify integration parameters and options for different integration scenarios.