envolution [1] filed a request to merge llama.cpp-cuda-f16 [2] into llama.cpp-cuda [3]: upstream now detects f16 mode automatically, negating the need for this package [1] https://aur.archlinux.org/account/envolution/ [2] https://aur.archlinux.org/pkgbase/llama.cpp-cuda-f16/ [3] https://aur.archlinux.org/pkgbase/llama.cpp-cuda/