b8730

The open-source project llama.cpp, maintained by ggml-org, has released a new update tagged b8730. This commit, authored by github-actions and cryptographically signed, focuses on a critical fix for the tokenizer used with Google's Gemma 4 model. Dubbed 'YATF (Yet Another Tokenizer Fix)', the patch addresses a specific edge case and, importantly, includes accompanying tests to ensure the fix is robust and to prevent regression in future updates. The release also removes an unnecessary hash from an update script, streamlining the development process.

The b8730 release is notable for its extensive list of pre-compiled binaries, making it easier for developers to deploy optimized versions of llama.cpp across diverse hardware ecosystems. The supported platforms now include macOS for both Apple Silicon (with optional KleidiAI acceleration) and Intel, various Linux configurations (supporting CPU, Vulkan, ROCm 7.2 for AMD GPUs, and Intel's OpenVINO), and Windows with support for CPU, CUDA 12.4/13.1, Vulkan, SYCL, and HIP. It also includes builds for Huawei's openEuler OS on both x86 and aarch64 architectures, targeting their Ascend AI processors (310p and 910b). This broad compatibility underscores the project's commitment to running large language models efficiently on consumer-grade and specialized hardware.