llamafile
Mozilla AIllamafile packages a large language model and the software to run it into a single cross-platform executable that runs without installation on macOS, Linux, BSD, and Windows. It combines llama.cpp with Cosmopolitan Libc to produce one file that works across operating systems and CPU architectures. Originally a Mozilla Builders project, it is now maintained by Mozilla.ai and ships a companion single-file speech-to-text tool, whisperfile.
Openness
5 high confidence- license
- Apache-2.0(OSI)+MIT(OSI, for llama.cpp/whisper.cpp deltas)
- source
- public(mozilla-ai/llamafile)
- core features withheld
- no
llamafile is licensed under Apache-2.0, with MIT covering its llama.cpp and whisper.cpp deltas. The source is public and the core carries no proprietary tier.
- https://github.com/mozilla-ai/llamafile recorded 2026-08-09
repo metadata: "isPrivate":false,"isArchived":false,"visibilityLabel":"Public"; License badge in README reads Apache 2.0; no pricing/paid-tier link anywhere on the repo page or in the documentation links listed
- https://raw.githubusercontent.com/mozilla-ai/llamafile/main/LICENSE recorded 2026-08-09
The Apache 2.0 License / Copyright 2023 Mozilla Foundation / Licensed under the Apache License, Version 2.0
- https://raw.githubusercontent.com/mozilla-ai/llamafile/main/README.md recorded 2026-08-09
While the llamafile project is Apache 2.0-licensed, our changes to llama.cpp and whisper.cpp are licensed under MIT (just like the projects themselves) so as to remain compatible and upstreamable in the future
Adoption
3 low confidencellamafile ships as single-file binaries rather than a package, so there is no download or pull count to measure adoption. GitHub stars are the only available signal, and because a star is not a use, they show attention to a popular local-inference tool without confirming how much it is actually run.
- https://github.com/mozilla-ai/llamafile recorded 2026-08-09
"stargazerCount":25523,"watcherCount":225,"forksCount":1544
Capability
3 medium confidenceOptimized for portability and zero-install local inference rather than datacenter serving throughput. GPU acceleration covers Apple Metal, NVIDIA, AMD and Vulkan, with the AMD and Windows paths documented as best-effort, which places it mid-tier among inference engines.
- https://docs.mozilla.ai/llamafile/reference/support recorded 2026-09-29
The Supported Systems page says "llamafile ships GPU acceleration for Apple Metal, NVIDIA, AMD, and Vulkan" and that "treat the AMD and Windows paths in particular as best-effort".
- https://raw.githubusercontent.com/mozilla-ai/llamafile/main/README_0.10.0.md recorded 2026-09-29
The 0.10.0 notes (a historical snapshot, superseded by the Supported Systems page) list "added Metal support: GPU on MacOS ARM64", "added TUI support", the llama.cpp server via the "--server" parameter, and, as then missing, "GPU support for Windows (and for whisperfile)".
- https://raw.githubusercontent.com/mozilla-ai/llamafile/main/README.md recorded 2026-09-29
The README says the project combines llama.cpp with Cosmopolitan Libc into one framework that collapses "all the complexity of LLMs down to" a single-file executable that runs "locally on most operating systems and CPU architectures, with no installation".
- https://raw.githubusercontent.com/mozilla-ai/llamafile/main/README.md recorded 2026-09-29
The README also describes "a single-file speech-to-text tool built on" whisper.cpp and says larger models can run with "external weights/models(GGUF)".
Verified 2026-09-27, except some scores, still awaiting re-confirmation