llama.cpp Reaches 1.0 with a Stable API and Inference Updates

The open-source inference project’s v1.0 release removes 35 deprecated functions and adds server hardening against SSRF and DNS rebinding, FP8 KV caches, and host-memory MoE caching. The related ggml v1.0.0 release adds lower-level engine updates, including API cleanup, scheduling, precision controls, RPC v9, and backend optimizations; migration impact and real-world performance remain to be validated.