llama.cpp releases version 1.0.0 with new versioning rules
November 16, 2026 · 3 min read
The development team for llama.cpp released version 1.0.0 on November 13, 2026. This release marks the first major version of the C and C++ library used to run large language model inference. The accompanying tensor library, ggml, also reached version 1.0.0 at the same time. The team announced that future incompatible changes to the public application programming interface and application binary interface will trigger a major version update.
llama.cpp is an open source project that supports a wide range of hardware, from personal computers to enterprise servers. At the time of this announcement, the project supports more than 150 model architectures, including derived forms. The development takes place under the ggml-org organization on GitHub. The team from the company ggml.ai, which led the development, announced its participation in Hugging Face in February 2026. The team stated that it would continue development and maintenance with long-term support from Hugging Face, while the community would continue to decide on technology and design autonomously.
The versioning policy applies to libllama, the library for large language model inference, and libggml, which supports its calculations. When a library update breaks application programming interface or application binary interface compatibility, applications that embed these libraries may require code modifications or rebuilding. The team outlined a rule for updating version numbers in the format of major, minor, and patch to communicate compatibility changes to developers.
In version 1.0.0, 35 deprecated functions, including llama_set_warmup, were removed from the public header llama.h. Code that calls these functions needs to be reviewed during the transition. The release also added a speech synthesis endpoint, /v1/audio/speech, to the inference server llama-server. The web user interface for the server now includes features to group chats into named groups and a Markdown editor to preview chat input content.
The development team explained that balancing the flexibility and stability of the application programming interface is important as the library becomes more embedded in other applications. About three and a half years have passed since development began. The team stated that it prioritizes maintaining code quality, deciding on design, and understanding implementation details while integrating large language models into the development process. The team is also investing in a continuous integration infrastructure to verify code changes. It is expanding environments for automatic code verification on major hardware, using Hugging Face services and receiving support from hardware vendors.
Georgi Gerganov, who started llama.cpp, announced the release on X, stating that a solid foundation for the future of local artificial intelligence has been built. He expressed expectations for what developers will create using the two libraries. Merve, who co-authored the official announcement article, also posted on X about the team's policy to maintain the robustness and lightness of the library while continuing to support all platforms and models. Clement Delangue, the chief executive officer of Hugging Face, stated that the company would develop openly with the community. He explained that enabling artificial intelligence to run on users' devices helps avoid the concentration of artificial intelligence in a few companies. Julien Chaumond, the chief technology officer of Hugging Face, stated that local artificial intelligence, privacy, and the ability to manage artificial intelligence oneself will become increasingly important.