Skip to content

An editorial team and a community building with AI

vibecoding.incJoin Discord ↗
← All articles

News · ggml-org (llama.cpp) · 2 min read

Event date: · Published:

llama.cpp and ggml reach their first major v1.0.0 release

ggml-org announced on November 13, 2026 that llama.cpp and its ggml tensor library reached version 1.0.0, the project's first major release in 3.5 years, with formal semantic versioning and support across five operating systems.

ggml-org, the team behind the llama.cpp inference engine, announced on November 13, 2026 that both libllama, the project's LLM inference library, and libggml, its underlying tensor library, reached version 1.0.0 simultaneously. The announcement came via a llama.cpp v1.0.0 blog post on Hugging Face, marking the first major-version release in the project's 3.5-year history.

What changed in v1.0.0

  • libllama and libggml reach v1.0.0 at the same time, the project's first major-version release after 3.5 years of development
  • The project formally adopts semantic versioning: MAJOR releases mark breaking C API changes, MINOR releases add backward-compatible features or new model and backend support, and PATCH releases fix bugs
  • More than 150 model architecture variants are supported, along with mainstream speculative-decoding methods
  • Builds are distributed as pre-release artifacts, Docker images, Debian packages, Homebrew formulas, conda-forge packages, MacPorts ports, Winget packages, and an XCFramework
  • The engine runs on Linux, macOS, Windows, Android, and iOS
  • Hardware backend partners credited in the release include NVIDIA, AMD, Qualcomm, Arm, IBM, and Intel

What it means for people who build with AI

llama.cpp underpins a large share of local and self-hosted LLM inference tooling, from Ollama-style wrappers to mobile apps and embedded deployments. A stable v1.0 release with a formal API and ABI versioning guarantee lets downstream projects upgrade with predictable compatibility instead of tracking a fast-moving, unversioned codebase.

Availability and infrastructure

The project's continuous integration now runs on Hugging Face's HF Jobs and HF Buckets services, reflecting the scale of builds needed to cover five operating systems and more than half a dozen package managers.

Sources

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.

llama.cpp and ggml reach their first major v1.0.0 release | vibecoding.inc