llama.cpp ships v1.0.0: major version numbers now flag API and ABI breaks, giving downstream apps a lockable interface boundary

llama.cpp released libraries v1.0.0 and established the major version number as the signal for API and ABI incompatibility, so downstream developers relying on its library interfaces can judge upgrade risk. About 3.5 years after the project began, many downstream applications already depend on its work, while interface stability had not been promised through a version-numbering rule. The release note says that as adoption grows the project will be stricter about making these boundaries visible to downstream consumers, and that the rule applies to all ggml-org projects using semantic versioning.

Beyond version numbering, execution capability still rests on ggml backends: each backend can be compiled as a dynamic library and optionally loaded at runtime. ggml is designed to run on virtually any existing hardware by implementing dedicated backends for different hardware. The release note says llama.cpp currently supports more than 150 architecture variants and all mainstream speculative decoding methods, with hardware support from NVIDIA, AMD, Qualcomm, Arm, IBM and Intel.

This item addresses downstream developers and packaging maintainers who depend on llama.cpp library interfaces, and describes version rules and current supported scope rather than specific performance figures.

The release note gives no performance numbers for individual backends or architecture variants, and does not state the concrete migration impact of the new version rule on existing downstream applications.