AI & Tech··1 sources verified
llama.cpp and ggml Ship v1.0.0 With a Major-Version Rule for C API Breaks
Summarised by Relevant News AI · Read time: 3 min
The llama.cpp and ggml projects tagged v1.0.0 on November 13, committing that any breaking change to their public C API will require a major-version bump. The release itself removes 35 deprecated functions, along with the llama_set_warmup function and the --defrag-thold option. It also adds FP8 KV caches and a host-memory expert cache for MoE models.
Why it matters: Developers who embed llama.cpp through rolling builds should test their integrations against removed functions before upgrading, and the new versioning policy gives them a clearer signal for future breaking changes.
All sources