DLA-future implements efficient multicore and GPU eigenvalue solvers, designed around C++'s std::execution concurrency proposal (P2300) as implemented in pika. DLA-Future takes advantage of asynchronous task-based programming, and it is designed to exploit modern heterogeneous architectures. DLA-Future also provides a C API, and a Fortran API shipped separately by DLA-Future-Fortran. These APIs allow to use DLA-Future as a drop-in replacement of ScaLAPACK eigensolvers, and have been integrated in popular electronic structure codes: CP2K and SIRIUS. DLA-Future-Fortran is also integrated in ELSI, a unified interface for electronic structure codes into high-performance eigensolvers, and therefore usable in SIESTA, and FHI-aims. We present weak scaling results for DLA-Future, up to matrix sizes of 500k x 500k, and results for realistic and large-scale electronic structure calculations with popular codes in which DLA-Future is integrated: CP2K and SIRIUS. Results are presented for modern hardware (NVIDIA GH200, AMD MI250x) available on the ALPS research infrastructure at the Swiss National Supercomputing Center (CSCS).