llama.cpp b9838 released with regex-partial cleanup and broad platform support
New release removes unused regex-partial code, supports macOS, Linux, Windows, Android, and more
ggml-org's llama.cpp, the leading open-source framework for running large language models locally, has released version b9838. This commit, signed with GitHub's verified signature, primarily removes the unused "regex-partial" component (issue #25118). While a small cleanup, it reflects ongoing maintenance efforts to keep the codebase clean and efficient. The release also bundles UI assets version 27, hinting at potential frontend improvements.
The build matrix for b9838 is extensive, covering macOS (Apple Silicon arm64 with and without KleidiAI, Intel x64, and iOS XCFramework), Linux (x64 CPU, arm64 CPU, s390x, x64 with Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16), Android (arm64 CPU), and Windows (x64 CPU, arm64 CPU, arm64 with OpenCL Adreno, x64 with CUDA 12/13, Vulkan, OpenVINO, SYCL, HIP). Notably, openEuler builds are disabled for this release, and some targets are marked as DISABLED. This breadth underscores llama.cpp's commitment to running AI inference on almost any device, from low-end hardware to high-end GPUs. For the AI community, such incremental updates ensure reliability and performance without introducing breaking changes.
- Removed unused regex-partial code to streamline the codebase
- Supports a vast array of platforms: Apple Silicon, Linux, Windows, Android, and more
- Includes UI assets version 27, suggesting interface improvements
Why It Matters
Keeps llama.cpp lean and compatible, enabling cost-effective LLM inference across diverse hardware.