On this page · 6 sections
  1. Major Version Numbers Now Signal Breaking C API Changes
  2. Which Version Bump Covers Which Change
  3. Projects That Link libllama and libggml
  4. Where the v1.0.0 Packages Are Distributed
  5. Other Items in the Announcement
  6. Sources

TL;DR

llama.cpp and ggml have released libraries v1.0.0. From now on, both projects will reflect API and ABI incompatibilities with major-version bumps, and PATCH is reserved for bug fixes with no API change, according to the Hugging Face announcement.

  • llama.cpp v1.0.0 ships as libllama, with public headers in llama.cpp/include/.
  • ggml v1.0.0 ships as libggml, with public headers in ggml/include/.
  • A breaking change to the public C API in ./include carries a MAJOR bump.
  • The same semantic version rules apply to all ggml-org projects that use semantic versioning.

Major Version Numbers Now Signal Breaking C API Changes #

The major version number becomes the signal for API and ABI incompatibilities in llama.cpp and ggml. The announcement gives the reason: every new integration adds developers who depend on the library interfaces, so a flexible but stable API is paramount. It also says the project will be stricter about making those boundaries visible to downstream consumers as adoption grows.

Moving forward, both llama.cpp and ggml will reflect API and ABI incompatibilities with major-version bumps.

— Hugging Face

The policy sentence speaks of API and ABI incompatibilities in general, while the rules table ties MAJOR specifically to a breaking change to the public C API in ./include. Readers should use that table as the exact scope when they read a version number.

Which Version Bump Covers Which Change #

The announcement assigns each change type a single bump. A breaking change to the public C API is the only change type listed under MAJOR, while API additions fall under MINOR. A new function in the headers therefore falls under MINOR, not MAJOR, by the table's own wording.

Change typeBump
Breaking change to the public C API in ./includeMAJOR
Backward-compatible feature, backend updates, model support, or API additionMINOR
Bug fix with no API changePATCH

Applications that build against the libraries depend on the public headers named in the announcement: llama.cpp/include/ for libllama and ggml/include/ for libggml. The MAJOR rule covers breaking changes to the public C API in ./include, so these are the directories where a major bump signals a break.

Where the v1.0.0 Packages Are Distributed #

The announcement's Distribution section says the project looks forward to expanding how ggml-org software reaches developers and end-users, and it links a discussion on adding ggml-org projects to packaging and release pipelines. The channels below are the ones it lists, with the platforms each covers.

ChannelPlatforms
Pre-release artifactsLinux, macOS, Windows, Android, iOS
DockerLinux
DebianLinux
HomebrewmacOS, Linux
conda-forgeWindows, macOS, Linux
MacPortsmacOS
WingetWindows
XCFrameworkiOS, macOS

Other Items in the Announcement #

The announcement also covers the backend and model coverage and the CI work. ggml backends can be compiled as dynamic libraries and optionally loaded at runtime, and llama.cpp supports more than 150 architecture variants as well as all mainstream speculative decoding methods. HF Buckets and HF Jobs are used to scale CI capacity, and self-hosted runners are being added across major hardware platforms, with hardware support from NVIDIA, AMD, Qualcomm, Arm, IBM and Intel.

The announcement names no security fix for v1.0.0.

The Sentinel’s Call

Projects that link libllama or libggml should check the headers in llama.cpp/include/ and ggml/include/ first, because a breaking change to the public C API in ./include is what carries a MAJOR bump. The announcement names no security fix, so nothing else in it ranks above that header check. The remaining items in the announcement do not change what that check covers.

Our judgement, drawn only from the facts and sources above.

Sources

  1. llama.cpp v1.0.0 : A stronger foundation for the local AI ecosystem huggingface.co · Nov 13, 2026

Next reading

Edited and approved by Jarosław Krzemiński. Spotted a mistake? Report a correction or read the corrections policy.