Llama.cpp version b9770 fixes remote preset handling and adds a test for it. The release includes binaries for macOS, Linux, Android, Windows, and openEuler, supporting various architectures and hardware accelerators like Vulkan, CUDA, OpenVINO, and SYCL.