llamafile Download: Run an AI Model From a Single File
llamafile 0.10.6 · for Windows, macOS and Linux
Developer: Mozilla.ai (opens in a new tab)
- Windows
- macOS
- Linux
The download link points to the program’s official source; the file is not hosted on this site. Link address: github.com
Link not working? Let us know
llamafile is a Mozilla.ai project that packs an open language model and llama.cpp into one executable file, so you can run AI locally without installing anything. The latest version is 0.10.6, released on September 15, 2026; it runs on Windows, macOS, Linux and BSD and is licensed under Apache 2.0.
Most local AI tools ask you to install an app and then download a model. llamafile turns that into a single step: the model and the program that runs it live inside one file. Download it, make it executable and run it.
What does llamafile do?
- One file, no install: It combines llama.cpp with Cosmopolitan Libc, so the same file runs on most operating systems and CPU architectures.
- Many systems: llamafiles run on macOS, Windows, Linux and the BSDs, on both AMD64 and ARM64 processors.
- Bring your own model: Besides ready-made llamafiles, the plain llamafile program can load external GGUF model files.
- Speech too: The bundled whisperfile tool does speech-to-text and translation in a single file, built on whisper.cpp.
Who should try it?
It suits anyone who wants to try a local model without setting up an app, and people who want to hand a working AI model to someone else as one file.
Running your first llamafile
There is nothing to install. The official quick start uses a small Qwen3.5 0.8B model, chosen because it is the smallest one and most likely to work on any computer:
curl -LO https://huggingface.co/mozilla-ai/llamafile_0.10/resolve/main/Qwen3.5-0.8B-Q8_0.llamafile
chmod +x Qwen3.5-0.8B-Q8_0.llamafile
./Qwen3.5-0.8B-Q8_0.llamafile
On Windows, rename the file so it ends in .exe and run it. Windows only runs executables smaller than 4 GB, so for bigger models get the plain llamafile program from the GitHub releases and load the GGUF weights as a separate file.
Version 0.10 introduced a new build system that tracks recent llama.cpp more closely. If you miss a feature from the older “classic” releases, those remain available on the releases page.
What hardware do you need?
- CPU: A 64-bit AMD64 or ARM64 processor.
- Memory and disk: These depend on the model. The 0.8B example is the lightest; larger, more capable models need more RAM and benefit from a GPU.
Licensing
llamafile is open source under the Apache 2.0 license; its changes to llama.cpp and whisper.cpp are MIT-licensed. Each model has its own license.
Pros
- Nothing to install: one file runs the model
- The same file works on several operating systems
- Can also load separate GGUF model files
Cons
- Windows cannot run single files over 4 GB
- Fewer conveniences than a full desktop chat app
Frequently Asked Questions
Do I need to install anything to use llamafile?
No. A llamafile is a single executable that contains the model and the program. On macOS, Linux and BSD you make it executable with chmod +x; on Windows you add .exe to the name.
Why does a large llamafile not run on Windows?
Windows does not run executables larger than 4 GB. Use the plain llamafile program and load the model weights as a separate GGUF file instead.
llamafileOpen source
Download (opens in a new tab)
