Skip to content
worth noting Open-source

llama.cpp and ggml reach version 1.0.0 with an API/ABI stability guarantee

clearly official source

The llama.cpp and ggml projects have released version 1.0.0, introducing semantic versioning and a formal guarantee that breaking changes to the API/ABI will carry a major version increment. Available through eight distribution channels, it supports more than 150 model architectures.

The llama.cpp project and its underlying tensor library ggml have released version 1.0.0, introducing formal semantic versioning. According to the developers, from now on every incompatible change to the public C API (in the include directory) will be signaled by a major version increment (MAJOR), while backward-compatible additions such as support for additional models or backends will carry only a MINOR version increment, and bug fixes without API changes will increment the PATCH version. This guarantee applies to the libllama (llama.cpp) and libggml (ggml) libraries.

The release comes after 3.5 years of development, during which a number of downstream applications dependent on its interfaces have emerged around the project. According to the developers, as adoption grows the team will be stricter about keeping the library's API boundaries visible and predictable for downstream consumers.

The project is now available through eight distribution channels for various platforms (Linux, macOS, Windows, Android, iOS) — including Docker, Debian packages, Homebrew, conda-forge, MacPorts, Winget, and XCFramework. For its CI infrastructure, the project uses Hugging Face's HF Buckets and HF Jobs technologies and draws on hardware support from self-hosted runners provided by NVIDIA, AMD, Qualcomm, Arm, IBM, and Intel. llama.cpp currently supports more than 150 model architectures and the main speculative decoding methods.

What changed

Why it matters

For developers building local AI tools on top of llama.cpp, this is a signal that the library is moving from development to production status — breaking changes will now be predictably signaled by a major version increment, enabling safer long-term update planning and reducing the risk that a minor library upgrade breaks an existing integration.

Two audiences, two different impacts

What this means

01

For individuals

Developers building their own tools on top of the libllama and libggml libraries can now rely on a stable public API and ABI between minor versions, which simplifies maintenance of code that depends on these libraries.

What to do Developers using llama.cpp/ggml can check whether their packaging pipeline matches the newly introduced semantic versioning (MAJOR/MINOR/PATCH).
More practical updates →
02

For a business

Companies that integrate llama.cpp or ggml into their own products now have, according to the developers, a formal guarantee that breaking changes to the API/ABI will only be signaled by a major version increment — this reduces the risk of unexpected outages when updating dependencies and makes it easier to plan release cycles.

Development
What to decide Companies building products on llama.cpp or ggml may want to review their library update process in light of the new semantic versioning commitment (major bump = breaking change).
More business impacts →
C/C++ ggml inference llama.cpp local AI open-source

Check the original

Event sources

clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.

1
Hugging Face Blog primary source · first detected llama.cpp v1.0.0 : A stronger foundation for the local AI ecosystem