Repolode
Submit
mozilla-ai

mozilla-ai/llamafile

Distributes and runs open LLMs as a single executable file that works across most operating systems and CPU architectures without installation.

Open-source alternative to:Ollama

View on GitHubWebsite
Stars
26,447
Forks
1,847
Open issues
216
Last commit
9 days ago

What is it?

llamafile packages a large language model and its inference runtime into one self-contained executable. It combines llama.cpp with Cosmopolitan Libc, so the same file can run locally on most operating systems and CPU architectures with no installation step. Users download a llamafile, mark it executable, and run it.

The project also includes whisperfile, a single-file speech-to-text tool built on whisper.cpp that handles transcription and translation of audio files on the same platforms. Starting with version 0.10.0, llamafile uses a new build system that tracks recent llama.cpp releases, so it supports newer models and features but may lack some options from earlier versions, which remain available on the releases page.

On Windows, executables larger than 4GB cannot run, so larger models must be loaded as external GGUF weights alongside the llamafile binary.

Key features

  • 01Packages model weights and runtime into one executable with no installation.
  • 02Runs on most operating systems and CPU architectures.
  • 03Built on llama.cpp for local LLM inference.
  • 04Includes whisperfile for audio transcription and translation.
  • 05Supports external GGUF weights for models too large for a single file.

Use cases

  • Sharing a working local LLM with colleagues or end users without setup steps.
  • Running a small open model offline on a laptop or desktop.
  • Transcribing or translating audio files locally with whisperfile.

Quick start

  1. 1

    Download an example llamafile model (Qwen3.5 0.8B)

    curl -LO https://huggingface.co/mozilla-ai/llamafile_0.10/resolve/main/Qwen3.5-0.8B-Q8_0.llamafile
  2. 2

    Make the file executable (macOS/Linux/BSD)

    chmod +x Qwen3.5-0.8B-Q8_0.llamafile
  3. 3

    Run the llamafile

    ./Qwen3.5-0.8B-Q8_0.llamafile

Things to know

  • Files larger than 4GB cannot run on Windows, so larger llamafiles need external GGUF weights.
  • Versions 0.10 and later use a new build system and may lack some features of the classic releases.

Topics

#llm#local-inference#gguf#llama-cpp#single-binary#speech-to-text#cross-platform#local-ai#local-llm#open-source-ai

FAQ

What is llamafile?
Distributes and runs open LLMs as a single executable file that works across most operating systems and CPU architectures without installation. llamafile packages a large language model and its inference runtime into one self-contained executable. It combines llama.cpp with Cosmopolitan Libc, so the same file can run locally on most operating systems and CPU architectures with no installation step. Users download a llamafile, mark it executable, and run it.
Is llamafile free to use? What license does it have?
llamafile is open source under the Other license. Check the license terms before using it in commercial products.
Is llamafile still maintained?
Actively maintained. The last commit to llamafile was on Oct 12, 2026.
What can I use llamafile for?
Sharing a working local LLM with colleagues or end users without setup steps. Running a small open model offline on a laptop or desktop. Transcribing or translating audio files locally with whisperfile.
How do I install llamafile?
Download an example llamafile model (Qwen3.5 0.8B): curl -LO https://huggingface.co/mozilla-ai/llamafile_0.10/resolve/main/Qwen3.5-0.8B-Q8_0.llamafile Make the file executable (macOS/Linux/BSD): chmod +x Qwen3.5-0.8B-Q8_0.llamafile Run the llamafile: ./Qwen3.5-0.8B-Q8_0.llamafile
Where can I get llamafile?
The source code and installation instructions are on GitHub: https://github.com/mozilla-ai/llamafile

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects