llama.cpp v9823 expands platform support with Windows OpenVINO CI
New release adds CUDA 13, OpenVINO, and ARM64 builds across Windows, Linux, and macOS.
ggml-org has released llama.cpp version b9823, a significant update to their widely-used C++ implementation of LLMs. This release, marked by commit 0275c0f and signed with a verified GPG key, introduces a new Windows OpenVINO CI check to ensure compatibility with Intel's OpenVINO toolkit. It also includes UI assets and a comprehensive set of prebuilt binaries targeting virtually every major hardware platform.
The release covers macOS Apple Silicon (with and without KleidiAI), macOS Intel, iOS XCFramework, and a wide array of Linux builds including CPU-only, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP64, and even s390x. Windows users get CPU, ARM64, OpenCL Adreno, CUDA 12.4/13.3 DLLs, Vulkan, OpenVINO, SYCL, and HIP variants. Android ARM64 and openEuler (x86 and ARM) with ACL Graph are also supported. This breadth ensures developers can deploy llama.cpp on everything from data center GPUs to edge devices.
- New Windows OpenVINO CI pipeline added for cross-platform validation.
- Supports CUDA 12.4 and 13.3 DLLs, plus ROCm 7.2, SYCL, and HIP.
- Includes Linux builds for s390x and openEuler with ACL Graph acceleration.
Why It Matters
Empowers developers to run local LLMs on any hardware, from cloud GPUs to ARM edge devices.