Local AI

local.ai is a free, open-source native app for downloading, verifying, and running AI models entirely offline, without needing a GPU. It manages model files with resumable downloads, checks their integrity using BLAKE3 and SHA256 digests, and runs inference on the CPU with automatic thread adaptation, so performance scales to the hardware available on a given machine. The app includes a local streaming inference server, which lets other applications call into it, for example through the window.ai standard, rather than only offering a built-in chat window. It supports several GGML quantization formats, giving users a choice between smaller, faster models and larger, more accurate ones. Built with a Rust backend, the app itself is reported to be under 10MB, and its source is licensed under GPLv3. local.ai is aimed at developers and researchers who want full local control over which models they run and how, rather than relying on a hosted API. It is available natively for Windows, macOS (Intel and Apple Silicon), and Linux. Because it depends entirely on CPU inference, larger models will run more slowly than on GPU-accelerated tools, which is a tradeoff for its no-GPU-required design.