llama.cpp and ggml released library version 1.0.0 and committed to marking API and ABI incompatibilities with major-version changes. Breaking public C API changes require a major bump, compatible features and model support use minor versions, and fixes without API changes use patch versions. llama.cpp supports more than 150 architecture variants, with broader hardware CI investment. The commitment makes compatibility boundaries explicit; it is not a promise to freeze interfaces permanently.