# llama.cpp v1.0.0 — first stable release, with semantic versioning

> The local inference engine behind much of open-model AI makes its first stable release and promises clear version rules.

- Org: ggml-org
- Date: 2026-11-13
- Category: tool, repo
- Source: https://huggingface.co/blog/ggml-org/llama-cpp-v1
- Canonical page: https://ai-tldr.dev/releases/llama-cpp-v1-0-0/

## In short

llama.cpp and ggml reach v1.0.0 after 3.5 years. Breaking API changes now bump the major version. The release removes 36 deprecated functions and adds an FP8 KV cache, Kimi-K3.1 support and a text-to-speech server endpoint.

## What it is

llama.cpp v1.0.0 is the first stable release of the C/C++ engine that runs open models locally, shipped together with ggml v1.0.0, the tensor library under it. The maintainers say the project, 3.5 years old and supporting more than 150 architecture variants, will now use semantic versioning across all ggml-org projects: a major bump for any breaking change to the public C API, a minor one for features and new models, a patch for fixes.

## How it works

To start the 1.x line clean, the release removes the deprecated public API, including llama_set_warmup and 35 other functions, and moves the NextN/speculative decoding API into llama.h with new MoE expert cache settings. Core gains an FP8 (E4M3) KV cache and can load PLE/engram weights from separate GGUF shards. llama-server adds an OpenAI-compatible /v1/audio/speech endpoint, cancels MCP tool calls when a client disconnects, and hardens SSRF, CORS and DNS-rebinding checks.

## Why it matters

Apps, wrappers and language bindings that embed libllama have had to chase silent API changes between builds. With version numbers that now signal breakage, downstream projects can pin to 1.x and upgrade safely. The flip side is immediate: code that still calls any of the 36 removed functions will not compile against v1.0.0.

## Who it's for

local LLM users, developers embedding libllama

## Try it

brew install llama.cpp

## Quick facts

- Project: llama.cpp and ggml (ggml-org)
- Version: v1.0.0 (build b12359)
- Released: 13 November 2026
- License: MIT
- Versioning: Semantic: MAJOR = breaking C API change
- Model support: 150+ architecture variants
- New in this build: FP8 (E4M3) KV cache, /v1/audio/speech endpoint

## FAQ

### Will llama.cpp v1.0.0 break my existing code?

llama.cpp v1.0.0 can break programs that link against libllama, because it removes the deprecated public API, including llama_set_warmup and 35 other deprecated functions. Code that still calls them must move to the current API before upgrading. ggml v1.0.0 also removes deprecated ops. People who only run the command-line tools or llama-server mostly need to read the release notes for changed options.

### What does semantic versioning mean for llama.cpp and ggml?

From v1.0.0, ggml-org projects follow semantic versioning. A MAJOR version bump means a breaking change to the public C API, a MINOR bump adds backward-compatible features or model support, and a PATCH bump fixes bugs without changing the API. The maintainers say they will be stricter about making API and ABI boundaries visible to downstream users.

### Which new models does llama.cpp v1.0.0 support?

The llama.cpp v1.0.0 release notes add text support for Kimi-K3.1-Pro, Ripple-65B-Preview, xing4_0 and the drama-3 text-to-speech model, plus vision support for MageVL-Cowork and Kimi-K3.1. These join more than 150 architecture variants the project already runs from GGUF files on consumer machines and servers.

### How do I install llama.cpp v1.0.0?

llama.cpp v1.0.0 ships through eight channels, according to the ggml-org announcement: pre-release artifacts on GitHub, Docker, Debian, Homebrew, conda-forge, MacPorts, Winget and an XCFramework for Apple platforms. The project is MIT-licensed, so it can be used in commercial products, and ggml backends can be built as dynamic libraries that load at runtime.

## Links

- [Announcement](https://huggingface.co/blog/ggml-org/llama-cpp-v1)
- [Release Notes](https://github.com/ggml-org/llama.cpp/releases/tag/v1.0.0)
- [ggml v1.0.0](https://github.com/ggml-org/ggml/releases/tag/v1.0.0)
- [Repository](https://github.com/ggml-org/llama.cpp)