llama.cpp b10038 adds official website, expands platform support
llama.cpp b10038 brings broader platform coverage and a new official site...
ggml-org's llama.cpp, the popular C/C++ library for running LLMs on consumer hardware, dropped version b10038 on July 16. The release is notable for its robust multi-platform support — from macOS Apple Silicon with KleidiAI acceleration to Linux with ROCm 7.2 and Windows with CUDA 12/13. The build matrix now covers CPU-only, Vulkan, OpenVINO, SYCL, and AMD HIP backends, plus Android arm64 and even openEuler Linux.
A key addition is the inclusion of an official website link (llama.app) in the release notes, signaling the project's maturation. While no new model support was announced, the breadth of platform builds (e.g., Windows arm64 with OpenCL Adreno, or Linux s390x) makes this a maintenance release that improves accessibility. For developers building local AI applications, b10038 ensures they can deploy on more architectures out of the box, reducing compilation friction.
- llama.cpp b10038 releases July 16 with builds for macOS, Linux, Windows, Android, iOS, and openEuler
- New official website (llama.app) added to release notes, aiding user discovery
- Includes specialized builds: KleidiAI for Apple Silicon, ROCm 7.2, CUDA 12/13, Vulkan, OpenVINO, SYCL, and OpenCL Adreno
Why It Matters
llama.cpp's latest update broadens local LLM deployment across nearly every platform, empowering developers with out-of-the-box builds.