The ggml-org team has released llama.cpp b11867, whose release notes mention a refactor of the server entry point. The release page provides builds for macOS, iOS and Linux, including CPU, Vulkan and CUDA options.
The available notes do not describe what changes for server users or whether the refactor affects performance or compatibility.
Why it matters
Developers choosing a llama.cpp build can find downloads for several platforms and hardware types, but the release notes provided do not say whether the server refactor changes setup or behavior.
