llamafile lets you distribute and run LLMs with a single file. (announcement blog post)

Our goal is to make open source large language models much more accessible to both developers and end users. We're doing that by combining llama.cpp with Cosmopolitan Libc into one framework that collapses all the complexity of LLMs down to a single-file executable (called a "llamafile") that runs locally on most computers, with no installation.

The easiest way to try it for yourself is to download our example llamafile for the LLaVA model (license: LLaMA 2, OpenAI). LLaVA is a new LLM that can do more than just chat; you can also upload images and ask it questions about them. With llamafile, this all happens locally; no data ever leaves your computer.

1. Download llava-v1.5-7b-q4-server.llamafile (3.97 GB).

2. Open your computer's terminal.

3. If you're using macOS, Linux, or BSD, you'll need to grant permission for your computer to execute this new file. (You only need to do this once.)

chmod +x llava-v1.5-7b-q4-server.llamafile

4. If you're on Windows, rename the file by adding ".exe" on the end.

5. Run the llamafile. e.g.:

./llava-v1.5-7b-q4-server.llamafile

6. Your browser should open automatically and display a chat interface. (If it doesn't, just open your browser and point it at https://localhost:8080.)

7. When you're done chatting, return to your terminal and hit Control-C to shut down llamafile.

Having trouble? See the "Gotchas" section below.

We also provide example llamafiles for two other models, so you can easily try out llamafile with different kinds of LLMs.

"Server llamafiles" work just like the LLaVA example above: you simply run them from your terminal and then access the chat UI in your web browser at https://localhost:8080.

"Command-line llamafiles" run entirely inside your terminal and operate just like llama.cpp's "main" function. This means you have to provide some command-line parameters, just like with llama.cpp.

Here is an example for the Mistral command-line llamafile:

./mistral-7b-instruct-v0.1-Q4_K_M-main.llamafile --temp 0.7 -r '\n' -p '### Instruction: Write a story about llamas\n### Response:\n'

And here is an example for WizardCoder-Python command-line llamafile:

./wizardcoder-python-13b-main.llamafile --temp 0 -r '\n' -p '\nvoid *memcpy_sse2(char *dst, const char *src, size_t size) {\n'

As before, macOS, Linux, and BSD users will need to use the "chmod" command to grant execution permissions to the file before running these llamafiles for the first time.

Unfortunately, Windows users cannot make use of these example llamafiles because Windows has a maximum executable file size of 4GB, and all of these examples exceed that size. (The LLaVA llamafile works on Windows because it is 30MB shy of the size limit.) But don't lose heart: llamafile allows you to use external weights; this is described later in this document.

Having trouble? See the "Gotchas" section below.

A llamafile is an executable LLM that you can run on your own computer. It contains the weights for a given open source LLM, as well as everything needed to actually run that model on your computer. There's nothing to install or configure (with a few caveats, discussed in subsequent sections of this document).

This is all accomplished by combining llama.cpp with Cosmopolitan Libc, which provides some useful capabilities:

1. llamafiles can run on multiple CPU microarchitectures. We added runtime dispatching to llama.cpp that lets new Intel systems use modern CPU features without trading away support for older computers.

2. llamafiles can run on multiple CPU architectures. We do that by concatenating AMD64 and ARM64 builds with a shell script that launches the appropriate one. Our file format is compatible with WIN32 and most UNIX shells. It's also able to be easily converted (by either you or your users) to the platform-native format, whenever required.

3. llamafiles can run on six OSes (macOS, Windows, Linux, FreeBSD, OpenBSD, and NetBSD). If you make your own llama files, you'll only need to build your code once, using a Linux-style toolchain. The GCC-based compiler we provide is itself an Actually Portable Executable, so you can build your software for all six OSes from the comfort of whichever one you prefer most for development.

4. The weights for an LLM can be embedded within the llamafile. We added support for PKZIP to the GGML library. This lets uncompressed weights be mapped directly into memory, similar to a self-extracting archive. It enables quantized weights distributed online to be prefixed with a compatible version of the llama.cpp software, thereby ensuring its originally observed behaviors can be reproduced indefinitely.

5. Finally, with the tools included in this project you can create your own llamafiles, using any compatible model weights you want. You can then distribute these llamafiles to other people, who can easily make use of them regardless of what kind of computer they have.

Even though our example llamafiles have the weights built-in, you don't have to use llamafile that way. Instead, you can download just the llamafile software (without any weights included) from our releases page. You can then use it alongside any external weights you may have on hand. External weights are particularly useful for Windows users because they enable you to work around Windows' 4GB executable file size limit.

For Windows users, here's an example for the Mistral LLM:

curl -o llamafile.exe https://github.com/Mozilla-Ocho/llamafile/releases/download/0.2.1/llamafile-server-0.2.1
curl -o mistral.gguf https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_M.gguf
.\llamafile.exe -m mistral.gguf

Here's the same example, but for macOS, Linux, and BSD users:

curl -L https://github.com/Mozilla-Ocho/llamafile/releases/download/0.2.1/llamafile-server-0.2.1 >llamafile
curl -L https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.1-GGUF/resolve/main/mistral-7b-instruct-v0.1.Q4_K_M.gguf >mistral.gguf
chmod +x llamafile
./llamafile -m mistral.gguf

On macOS with Apple Silicon you need to have Xcode installed for llamafile to be able to bootstrap itself.

If you use zsh and have trouble running llamafile, try saying sh -c ./llamafile. This is due to a bug that was fixed in zsh 5.9+. The same is the case for Python subprocess, old versions of Fish, etc.

On some Linux systems, you might get errors relating to run-detectors or WINE. This is due to binfmt_misc registrations. You can fix that by adding an additional registration for the APE file format llamafile uses:

sudo wget -O /usr/bin/ape https://cosmo.zip/pub/cosmos/bin/ape-$(uname -m).elf
sudo chmod +x /usr/bin/ape
sudo sh -c "echo ':APE:M::MZqFpD::/usr/bin/ape:' >/proc/sys/fs/binfmt_misc/register"
sudo sh -c "echo ':APE-jart:M::jartsr::/usr/bin/ape:' >/proc/sys/fs/binfmt_misc/register"

As mentioned above, on Windows you may need to rename your llamafile by adding .exe to the filename.

Also as mentioned above, Windows also has a maximum file size limit of 4GB for executables. The LLaVA server executable above is just 30MB shy of that limit, so it'll work on Windows, but with larger models like WizardCoder 13B, you need to store the weights in a separate file. An example is provided above; see "Using llamafile with external weights."

On WSL, it's recommended that the WIN32 interop feature be disabled:

sudo sh -c "echo -1 > /proc/sys/fs/binfmt_misc/WSLInterop"

On any platform, if your llamafile process is immediately killed, check if you have CrowdStrike and then ask to be whitelisted.

On Apple Silicon, everything should just work if Xcode is installed.

On Linux, Nvidia cuBLAS GPU support will be compiled on the fly if (1) you have the cc compiler installed, (2) you pass the --n-gpu-layers 35 flag (or whatever value is appropriate) to enable GPU, and (3) the CUDA developer toolkit is installed on your machine and the nvcc compiler is on your path.

On Windows, that usually means you need to open up the MSVC x64 native command prompt and run llamafile there, for the first invocation, so it can build a DLL with native GPU support. After that, $CUDA_PATH/bin still usually needs to be on the $PATH so the GGML DLL can find its other CUDA dependencies.

In the event that GPU support couldn't be compiled and dynamically linked on the fly for any reason, llamafile will fall back to CPU inference.

Here's how to build llamafile from source. First, you need the cosmocc toolchain, which is a fat portable binary version of GCC. Here's how you can download the latest release and add it to your path.

mkdir -p cosmocc
cd cosmocc
curl -L https://github.com/jart/cosmopolitan/releases/download/3.1.3/cosmocc-3.1.3.zip >cosmocc.zip
unzip cosmocc.zip
cd ..
export PATH="$PWD/cosmocc/bin:$PATH"

You can now build the llamafile repository by running make:

make -j8

Here's an example of how to generate code for a libc function using the llama.cpp command line interface, utilizing WizardCoder-Python-13B weights:

make -j8 o//llama.cpp/main/main
o//llama.cpp/main/main \
  -m ~/weights/wizardcoder-python-13b-v1.0.Q8_0.gguf \
  --temp 0 \
  -r $'```\n' \
  -p $'```c\nvoid *memcpy_sse2(char *dst, const char *src, size_t size) {\n'

Here's a similar example that instead utilizes Mistral-7B-Instruct weights:

make -j8 o//llama.cpp/main/main
o//llama.cpp/main/main \
  -m ~/weights/mistral-7b-instruct-v0.1.Q4_K_M.gguf \
  --temp 0.7 \
  -r $'\n' \
  -p $'### Instruction: Write a story about llamas\n### Response:\n'

Here's an example of how to run llama.cpp's built-in HTTP server in such a way that the weights are embedded inside the executable. This example uses LLaVA v1.5-7B, a multimodal LLM that works with llama.cpp's recently-added support for image inputs.

make -j8

o//llamafile/zipalign -j0 \
  o//llama.cpp/server/server \
  ~/weights/llava-v1.5-7b-Q8_0.gguf \
  ~/weights/llava-v1.5-7b-mmproj-Q8_0.gguf

o//llama.cpp/server/server \
  -m llava-v1.5-7b-Q8_0.gguf \
  --mmproj llava-v1.5-7b-mmproj-Q8_0.gguf \
  --host 0.0.0.0

The above command will launch a browser tab on your personal computer to display a web interface. It lets you chat with your LLM and upload images to it.

If you want to be able to just say:

./server

...and have it run the web server without having to specify arguments (for the paths you already know are in there), then you can add a special .args to the zip archive, which specifies the default arguments. In this case, we're going to try our luck with the normal zip command, which requires we temporarily rename the file. First, let's create the arguments file:

cat <<EOF >.args
-m
llava-v1.5-7b-Q8_0.gguf
--mmproj
llava-v1.5-7b-mmproj-Q8_0.gguf
--host
0.0.0.0
...
EOF

As we can see above, there's one argument per line. The ... argument optionally specifies where any additional CLI arguments passed by the user are to be inserted. Next, we'll add the argument file to the executable:

mv o//llama.cpp/server/server server.com
o//llamafile/zipalign -j0 server.com .args
mv server.com server
./server

Congratulations. You've just made your own LLM executable that's easy to share with your friends.

SYNOPSIS

  o//llamafile/zipalign ZIP FILE...

DESCRIPTION

  Adds aligned uncompressed files to PKZIP archive

  This tool is designed to concatenate gigabytes of LLM weights to an
  executable. This command goes 10x faster than `zip -j0`. Unlike zip
  you are not required to use the .com file extension for it to work.
  But most importantly, this tool has a flag that lets you insert zip
  files that are aligned on a specific boundary. The result is things
  like GPUs that have specific memory alignment requirements will now
  be able to perform math directly on the zip file's mmap()'d weights

FLAGS

  -h        help
  -N        nondeterministic mode
  -a INT    alignment (default 65536)
  -j        strip directory components
  -0        store uncompressed (currently default)

Here is a succinct overview of the tricks we used to create the fattest executable format ever. The long story short is llamafile is a shell script that launches itself and runs inference on embedded weights in milliseconds without needing to be copied or installed. What makes that possible is mmap(). Both the llama.cpp executable and the weights are concatenated onto the shell script. A tiny loader program is then extracted by the shell script, which maps the executable into memory. The llama.cpp executable then opens the shell script again as a file, and calls mmap() again to pull the weights into memory and make them directly accessible to both the CPU and GPU.

The trick to embedding weights inside llama.cpp executables is to ensure the local file is aligned on a page size boundary. That way, assuming the zip file is uncompressed, once it's mmap()'d into memory we can pass pointers directly to GPUs like Apple Metal, which require that data be page size aligned. Since no existing ZIP archiving tool has an alignment flag, we had to write about 400 lines of code to insert the ZIP files ourselves. However, once there, every existing ZIP program should be able to read them, provided they support ZIP64. This makes the weights much more easily accessible than they otherwise would have been, had we invented our own file format for concatenated files.

On Intel and AMD microprocessors, llama.cpp spends most of its time in the matmul quants, which are usually written thrice for SSSE3, AVX, and AVX2. llamafile pulls each of these functions out into a separate file that can be #includeed multiple times, with varying __attribute__((__target__("arch"))) function attributes. Then, a wrapper function is added which uses Cosmopolitan's X86_HAVE(FOO) feature to runtime dispatch to the appropriate implementation.

llamafile solves architecture portability by building llama.cpp twice: once for AMD64 and again for ARM64. It then wraps them with a shell script which has an MZ prefix. On Windows, it'll run as a native binary. On Linux, it'll extract a small 8kb executable called APE Loader to ${TMPDIR:-${HOME:-.}}/.ape that'll map the binary portions of the shell script into memory. It's possible to avoid this process by running the assimilate program that comes included with the cosmocc compiler. What the assimilate program does is turn the shell script executable into the host platform's native executable format. This guarantees a fallback path exists for traditional release processes when it's needed.

Cosmopolitan Libc uses static linking, since that's the only way to get the same executable to run on six OSes. This presents a challenge for llama.cpp, because it's not possible to statically link GPU support. The way we solve that is by checking if a compiler is installed on the host system. For Apple, that would be Xcode, and for other platforms, that would be nvcc. llama.cpp has a single file implementation of each GPU module, named ggml-metal.m (Objective C) and ggml-cuda.cu (Nvidia C). llamafile embeds those source files within the zip archive and asks the platform compiler to build them at runtime, targeting the native GPU microarchitecture. If it works, then it's linked with platform C library dlopen() implementation. See llamafile/cuda.c and llamafile/metal.c.

In order to use the platform-specific dlopen() function, we need to ask the platform-specific compiler to build a small executable that exposes these interfaces. On ELF platforms, Cosmopolitan Libc maps this helper executable into memory along with the platform's ELF interpreter. The platform C library then takes care of linking all the GPU libraries, and then runs the helper program which longjmp()'s back into Cosmopolitan. The executable program is now in a weird hybrid state where two separate C libraries exist which have different ABIs. For example, thread local storage works differently on each operating system, and programs will crash if the TLS register doesn't point to the appropriate memory. The way Cosmopolitan Libc solves that is by JITing a trampoline around each dlsym() import, which blocks signals using sigprocmask() and changes the TLS register using arch_prctl(). Under normal circumstances, aspecting each function call with four additional system calls would be prohibitively expensive, but for llama.cpp that cost is infinitesimal compared to the amount of compute used for LLM inference. Our technique has no noticeable slowdown. The major tradeoff is that, right now, you can't pass callback pointers to the dlopen()'d module. Only one such function needed to be removed from the llama.cpp codebase, which was an API intended for customizing logging. In the future, Cosmoplitan will just trampoline signal handlers and code morph the TLS instructions to avoid these tradeoffs entirely. See cosmopolitan/dlopen.c for further details.

The example llamafiles provided above should not be interpreted as endorsements or recommendations of specific models, licenses, or data sets on the part of Mozilla.

llamafile adds pledge() and SECCOMP sandboxing to llama.cpp. This is enabled by default. It can be turned off by passing the --unsecure flag. Sandboxing is currently only supported on Linux and OpenBSD on systems without GPUs; on other platforms it'll simply log a warning.

Our approach to security has these benefits:

1. After it starts up, your HTTP server isn't able to access the filesystem at all. This is good, since it means if someone discovers a bug in the llama.cpp server, then it's much less likely they'll be able to access sensitive information on your machine or make changes to its configuration. On Linux, we're able to sandbox things even further; the only networking related system call the HTTP server will allowed to use after starting up, is accept(). That further limits an attacker's ability to exfiltrate information, in the event that your HTTP server is compromised.

2. The main CLI command won't be able to access the network at all. This is enforced by the operating system kernel. It also won't be able to write to the file system. This keeps your computer safe in the event that a bug is ever discovered in the the GGUF file format that lets an attacker craft malicious weights files and post them online. The only exception to this rule is if you pass the --prompt-cache flag without also specifying --prompt-cache-ro. In that case, security currently needs to be weakened to allow cpath and wpath access, but network access will remain forbidden.

Therefore your llamafile is able to protect itself against the outside world, but that doesn't mean you're protected from llamafile. Sandboxing is self-imposed. If you obtained your llamafile from an untrusted source then its author could have simply modified it to not do that. In that case, you can run the untrusted llamafile inside another sandbox, such as a virtual machine, to make sure it behaves how you expect.

While the llamafile project is Apache 2.0-licensed, our changes to llama.cpp are licensed under MIT (just like the llama.cpp project itself) so as to remain compatible and upstreamable in the future, should that be desired.

The llamafile logo on this page was generated with the assistance of DALL·E 3.