llama.cpp Ships v1.0.0 With Semver for libllama and ggml

The ggml-org team at Hugging Face has shipped llama.cpp v1.0.0 and ggml v1.0.0, formalizing semantic versioning for the inference libraries after what the team's blog post describes as "3.5 years since its inception." From here on, the maintainers say, "both llama.cpp and ggml will reflect API and ABI incompatibilities with major-version bumps."

The rules are spelled out. A MAJOR bump means a breaking change to the public C API in `./include`. MINOR covers backward-compatible features, backend updates, and new model support. PATCH is reserved for bug fixes with no API change. "As adoption grows, we will be stricter about making those boundaries visible to downstream consumers," the team writes.

Coverage is broad. llama.cpp supports "more than 150 architecture variants as well as all mainstream speculative decoding methods," with pre-release artifacts for Linux, macOS, Windows, Android and iOS, plus packaging through Docker, Debian, Homebrew, conda-forge, MacPorts, Winget and an XCFramework. Hardware vendors contributing self-hosted CI runners include NVIDIA, AMD, Qualcomm, Arm, IBM and Intel, with scaling handled by "HF technologies such as HF Buckets and HF Jobs."

The release post names six Hugging Face authors, julien-c, ggerganov, clem, pcuenq, raunoryys and nisbenz, alongside "104 additional supporters" from the contributor community. "Maintaining a high quality codebase, staying in control of the architecture and understanding the implementation details will remain our top priorities," they write, pitching the result as "the most lightweight, versatile and sensible platform for building cross-platform and private AI products." It lands against a steady run of 138 inference stories tracked in our feed over the past 90 days.