mudler/localai is an open-source AI engine that runs LLM, vision, speech, image, and video models on a wide range of hardware and serves them through an API. The latest public release is v4.11.0. The provided record lists no items added or changed in this release, so beyond the version number, no changes can be confirmed.
The project is a self-hosted engine for users who want to operate models on their own infrastructure. According to the README, it can run models without a GPU and loads only the backends needed for a given request. It provides an OpenAI-, Anthropic-, and ElevenLabs-API-compatible interface across multiple backends. On-device and edge deployments, as well as distributed and P2P execution of backends, are also introduced as server features. Its built-in AI agent supports tool use, RAG (retrieval-augmented generation), MCP, and skills, and it also includes API key authentication, per-user quotas, and role-based access. The project describes itself as designed so that data does not leave the user's infrastructure.
The repository is written in Go and is licensed under MIT. However, the provided material contains no changelog or performance measurements for v4.11.0. Therefore, what improved in this version, or how well it performs on each piece of hardware, cannot be judged from this record alone. The only confirmable point is that it is an engine designed to serve multiple model types and backends through an API in a self-operated environment.