Hugging Face releases llama.cpp v1.0.0 enabling local AI inference across devices

Hugging Face’s engineering team released llama.cpp v1.0.0, a C/C++ inference library for multimodal large language models, together with ggml v1.0.0, the underlying tensor library Hugging Face Engineering Blog. The version bump introduces explicit major-version boundaries for API and ABI changes, signalling stricter compatibility rules for downstream developers.

Both libraries expose public headers (llama.cpp/include/ and ggml/include/) and are distributed as pre-release artifacts for Linux, macOS, Windows, Android and iOS, plus Docker images, Debian packages, Homebrew formulas, conda-forge builds, MacPorts ports, Winget packages and XCFrameworks. This breadth of channels aims to simplify integration into diverse packaging pipelines.

The project now supports more than 150 architecture variants and all mainstream speculative decoding methods, with dedicated backends that can be compiled as dynamic libraries and loaded at runtime. Hugging Face notes ongoing investment in CI infrastructure and hardware support from NVIDIA, AMD, Qualcomm, Arm, IBM and Intel to maintain correctness across the expanding hardware landscape.