mozilla-ai/llamafile

llamafile lets you package and run a large language model (an AI that can understand and generate text) as a single downloadable file, with no installation or setup required. The problem it solves is that running AI models locally normally involves installing multiple tools, dependencies, and libraries, a process that can be confusing and error-prone. llamafile collapses all of that into one self-contained executable that works on most major operating systems and CPU types: macOS, Linux, Windows, and BSD. The way it works is by combining two existing tools, llama.cpp, which handles running AI language models efficiently, and Cosmopolitan Libc, a special library that allows a single compiled program to run across different operating systems without modification. The resulting file contains both the software and the model weights, so you simply download it, mark it as executable, and run it. It also includes whisperfile, a companion tool for speech-to-text (turning spoken audio into written text), built on the same packaging approach. You would use this if you want to experiment with AI language models on your own computer without sending your data to a cloud service, or if you want to share a working AI model with someone else as a simple file download. Developers distributing AI-powered tools to non-technical users would find it especially useful. The project is built in C++ and is a Mozilla Builders initiative.