llama.cpp v1.0.0 lands with semver pledge, 150+ architectures

The 1.0.0 tag formalizes a relationship that had been informal for years. The team describes the milestone as arriving 3.5 years since llama.cpp's inception, and the release post pairs it with a promise: both llama.cpp and ggml will reflect API and ABI incompatibilities with major-version bumps. For a project whose audience increasingly builds products on top of it, that is a shift from moving-target codebase to versioned dependency.

The release also sketches the project's reach. Beyond more than 150 architecture variants and all mainstream speculative decoding methods, the post lists pre-release artifacts across five operating systems and packaging through seven channels, from Docker to an XCFramework. Hardware vendors contributing self-hosted CI runners — NVIDIA, AMD, Qualcomm, Arm, IBM and Intel — suggest the testing load is distributed across the very vendors whose chips the library targets, with scaling handled by HF Buckets and HF Jobs.

The maintainers frame their priorities as maintaining a high quality codebase, staying in control of the architecture and understanding the implementation details, pitching the result as the most lightweight, versatile and sensible platform for building cross-platform and private AI products. The release also credits six named Hugging Face authors and 104 additional supporters, and lands against a steady run of 138 inference stories tracked in the feed over the past 90 days — a reminder of how crowded the inference-tooling space has become.