Llamafiles

Llamafiles package a Large Language Model (LLM) and its runtime into a single, cross-platform executable, enabling local, no-installation AI.

Llamafiles radically simplify Large Language Model (LLM) deployment: they are single-file executables containing both the model weights and the necessary runtime. Developed through a collaboration between Mozilla and creator Justine Tunney, this technology combines the efficient `llama.cpp` inference engine with the `Cosmopolitan Libc` universal C library. This powerful pairing allows the file to run directly on six operating systems (Windows, macOS, Linux, and three BSD variants) without installation or complex dependencies. The result is a shift of AI from the cloud to local consumer hardware, ensuring data privacy and immediate, offline access to open models like LLaVA and Mistral.

https://mozilla-ai.github.io/llamafile