LocalAI is an open-source AI inference engine that lets users run language, vision, voice, image, and video models on their own hardware behind familiar APIs. It supports CPUs as well as NVIDIA, AMD, Intel, Apple Silicon, and Vulkan-capable hardware, and can load different backends such as llama.cpp, vLLM, Whisper, Stable Diffusion, and MLX only when needed.

The platform provides API compatibility with services such as OpenAI, Anthropic, and ElevenLabs, allowing many existing applications to switch to self-hosted models with minimal changes. It also includes authentication, quotas, role-based access, agents, RAG, MCP, and multi-user features. LocalAI is particularly useful for privacy-sensitive deployments, offline environments, local experimentation, and organizations that want one extensible API layer across multiple AI modalities without sending data to external providers.