What is LlamaChat and what can you do with it?

LlamaChat is a free, open-source macOS application for chatting with LLaMA-family language models entirely on your own Mac. You can use it to hold conversations with models such as LLaMA, Alpaca, GPT4All, and Vicuna without sending prompts to a cloud service. It requires macOS 13 and runs on both Intel processors and Apple Silicon. The app itself ships without any model files, so you must obtain and add a compatible model yourself before you can chat.

What you can do with LlamaChat

• Chat locally with supported models. The app is built for LLaMA, Alpaca, and GPT4All models, with Vicuna listed as coming soon. Alpaca, for example, is Stanford's 7B-parameter LLaMA model fine-tuned on instruction-following demonstrations, which is what gives it a chatbot-style response pattern rather than raw text completion.
• Import models in two formats. LlamaChat can import raw published PyTorch model checkpoints directly, or pre-converted .ggml model files. This means you can either bring a checkpoint you already have or use a model file that has already been converted.
• Keep everything on-device. Because the models run locally, your conversations stay on your Mac rather than being processed by a remote server.
• Inspect or contribute to the code. The app is fully open-source, powered by the open-source libraries llama.cpp and llama.swift, and the source is available on GitHub.

Requirements and installation

You can install LlamaChat in either of two ways:

1. Download the app from the project's site (version 1.2.0).
2. Install via Homebrew with brew install --cask llamachat.

After installing, you still need to add a model. Obtain a compatible LLaMA-family model file in accordance with its provider's terms, then import it into LlamaChat as either a raw PyTorch checkpoint or a .ggml file. Once the model is loaded, you can start a chat.

What to check before you commit

• Do you have a model? LlamaChat does not distribute any model files, proprietary or otherwise. You are responsible for obtaining and integrating model files under the terms set by their providers. If you don't already have a compatible model, factor that step into your decision.
• Is your Mac on macOS 13? The app requires macOS 13, so older systems are out.
• Do you need Vicuna? It is listed as coming soon rather than currently supported, so don't count on it yet.
• Do you want a GUI over a command-line setup? LlamaChat wraps the llama.cpp and llama.swift libraries in a Mac app, which suits people who want local chat without assembling the underlying tooling themselves.

LlamaChat is an independent application and is not affiliated with, endorsed by, or sponsored by Meta Platforms, Leland Stanford Junior University, or Nomic AI.