DAILY BRIEF / LOCAL INFERENCE
llama.cpp reaches v1.0.0 with semantic versioning.
What changed
The release removes 35 deprecated functions, promotes the NextN speculative API to llama.h, and adds FP8 KV caches and an MoE expert cache.
Why it matters
From now on, a MAJOR bump signals a breaking C API change. Code that binds to llama.h should be checked against the removed functions.
Scope and limitations
--defrag-thold is removed. The server now needs explicit opt-ins for unauthenticated tools and private media URLs.
Go to the source
A sourced technical brief published 18 Nov 2026. The implication is turingfold’s interpretation. We have not independently benchmarked this change.