llama.cpp b9202 ships with cmake fix and expanded platform builds

llama.cpp’s latest release, tag b9202 (commit e0de4c2), is now available, bringing a focused fix and broad platform coverage. The key change addresses a cmake issue where conversion scripts were incorrectly installed alongside the main library. This fix, referenced in pull request #23204, ensures a cleaner build process for users compiling from source, especially those deploying models in production or containerized environments. The release is signed with GitHub’s verified GPG key, adding an extra layer of trust for security-conscious developers.

Beyond the cmake fix, b9202 expands accessibility by providing precompiled binaries for nearly every major hardware configuration. macOS users get builds for Apple Silicon (with and without KleidiAI optimizations), Intel x64, and iOS XCFramework. Linux supports x64, arm64, and s390x CPUs, plus GPU backends including Vulkan, ROCm 7.2, OpenVINO, and SYCL (FP32 and FP16). Windows covers x64 and arm64 CPUs, CUDA 12/13, Vulkan, SYCL, and HIP. Android arm64, openEuler (x86 and aarch64 with ACL Graph) are also included. This near-universal coverage lets developers run state-of-the-art LLMs like Llama 3, Mistral, and Gemma on local hardware without manual compilation—critical for privacy, latency, and offline use cases.