llama.cpp v9173 expands platform support with CI fix
New release brings broader hardware compatibility and streamlined builds...
The open-source community behind llama.cpp pushed version b9173 on May 15, 2024, focusing on infrastructure reliability and platform compatibility. The primary change is a CI fix for release symlinks (commit 49d1701, signed with GitHub verified signature), ensuring that binary distribution links work correctly across all builds.
This release dramatically expands supported hardware. New builds include macOS Apple Silicon (both standard and KleidiAI-enabled), macOS Intel, iOS XCFramework, and Android arm64. Linux users get CPU, Vulkan, ROCm 7.2, OpenVINO, and SYCL variants for x64 and arm64. Windows adds CPU, arm64, CUDA 12/13, Vulkan, SYCL, and HIP. Even openEuler (310p and 910b with ACL Graph) gets builds. This means developers can deploy llama.cpp on practically any desktop, server, or mobile device.
- Fixes release symlinks in CI, ensuring stable distribution of binaries
- Adds macOS/Apple Silicon KleidiAI builds for optimized inference
- Expands support to 20+ build targets including Android, iOS, openEuler, and multiple GPU backends
Why It Matters
llama.cpp's universal platform support lets developers run local LLMs on any device, democratizing AI inference.