After three and a half years, llama.cpp has a 1.0. The inference engine that made running open models on ordinary hardware normal — written in C and C++, built on the ggml tensor library — shipped v1.0.0 on November 13 alongside ggml v1.0.0.
The headline is not a new feature but a promise. Both projects now follow semantic versioning: a breaking change to the public C API in `./include` means a major bump, backward-compatible features, backend updates, new model support or API additions mean a minor bump, and bug fixes mean a patch. The team frames it as a debt to downstream developers: "Every new integration means more developers depending on our library interfaces and therefore having a flexible but stable API is paramount."
The ecosystem the release speaks for is large and industrial. llama.cpp supports more than 150 model architecture variants and all mainstream speculative decoding methods, and its ggml backends can be compiled as dynamic libraries and optionally loaded at runtime across essentially any hardware. Distribution already covers Docker, Debian, Homebrew, conda-forge, MacPorts, Winget and an XCFramework for iOS and macOS, alongside pre-release artifacts for Linux, macOS, Windows, Android and iOS. Continuous integration runs on self-hosted runners with hardware from Nvidia, AMD, Qualcomm, Arm, IBM and Intel.
The practical consequence for anyone shipping local or private AI: upgrades become legible. If an update breaks your build, the version number should now say so.




