@atmorouter.dev: llama.cpp just hit v1.0.0, and the version number finally means something. On Nov 13, the llama.cpp team released llama.cpp v1.0.0 and ggml v1.0.0, 3.5 years after the project began. From now on, both libraries show API and ABI breaks with a major-version bump. The new rules for all ggml-org projects: • Major: a breaking change to the public C API • Minor: new features, backend updates, model support or API additions • Patch: a bug fix with no API change Before you upgrade: v1.0.0 removes llama_set_warmup, 35 deprecated functions and the defrag threshold option from llama.h. Code that still calls them will not compile. Also in 1.0: FP8 KV caches, a MoE expert cache, an OpenAI-compatible /v1/audio/speech endpoint in llama-server, security fixes (SSRF, CORS, DNS rebinding) and new models such as Kimi-K3.1-Pro and Ripple-65B-Preview. It supports 150+ model architecture variants and ships through 8 channels, including Docker, Homebrew, Debian and Winget. Sources: https://huggingface.co/blog/ggml-org/llama-cpp-v1 https://github.com/ggml-org/llama.cpp/releases/tag/v1.0.0 Run local where it fits and route the rest. AtmoRouter is the cheapest AI API gateway: one OpenAI-compatible endpoint for 100+ models, up to 95% off official price. atmorouter.dev #AI #LLM #llamacpp #LocalLLM #OpenSource #HuggingFace #AtmoRouter

atmorouter.dev

Open In TikTok:

Region: ID

Sunday 15 November 2026 00:05:00 GMT