llama.cpp and ggml have reached v1.0.0, ggml-org announced on November 13, 2026. Alongside the releases, the team is committing to major-version bumps when changes make their programming or binary interfaces incompatible. The version number is getting a job beyond looking official. Source
These are libraries for running AI, not a new language model: llama.cpp provides large language model inference in C/C++, while ggml is its underlying machine-learning tensor library. The announcement describes llama.cpp running multimodal models on hardware ranging from consumer machines to enterprise servers. Source
What the numbers are supposed to tell you
“Moving forward, both llama.cpp and ggml will reflect API and ABI incompatibilities with major-version bumps,” the team writes. The application programming interface (API) defines how developers use the library in source code; the application binary interface (ABI) governs how compiled programs interact with it. Source
For ggml-org projects using semantic versioning, the announcement lays out these rules:
- Major: A breaking change to the public C API in
./include.
- Minor: Backward-compatible updates, including features, backend updates, model support or API additions.
- Patch: A bug fix with no API change. Source
The team says applications already depend on its library interfaces, making a flexible but stable API increasingly important. Breaking changes can still happen; they're supposed to wear a much more obvious name tag. Source
The announcement also lists Homebrew, Debian, Docker, conda-forge and Winget among its distribution channels, and points developers to a separate packaging discussion. It doesn't confirm whether v1.0.0 is already available through each one. Source
Comments