Run Meta Llama 3 Locally on Apple Silicon Macs with Ollama & OpenWebUI

Looking for the simplest way to run the latest Meta Llama 3 on an Apple Silicon Mac? This guide walks you through installing Ollama, selecting a model, and setting up a user‑friendly UI with OpenWebUI.

What you need: Ollama

Ollama is a tool designed to run large language models such as Llama 3, Mistral, and Gemma directly on macOS. Get it from the official GitHub repository:

https://github.com/ollama/ollama

Install Ollama on macOS

After installation, open a new terminal session and verify the setup:

ollama -v

Choose your model

Depending on your Mac’s resources, you can run Meta Llama 3 8B or the larger 70B version. The 70B model requires substantial RAM (e.g., 40 GB); the 8B model needs about 4.7 GB. An M1 Max with 64 GB RAM handled the 70B model well.

Run the 8B model: ollama run llama3:8b Run the 70B model: ollama run llama3:70b Download time varies with your network speed.

Set up the user interface

After the model is ready, use OpenWebUI to interact with it. OpenWebUI runs in Docker, so install Docker first and then execute:

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Open your browser at http://localhost:3000/. Register a local account (no internet data is sent) and log in to see a ChatGPT‑like interface.

You can now select the Meta Llama 3 model (or any other open‑source LLM) and start testing.

Conclusion

Once configured, your Apple Silicon Mac becomes a powerful hub capable of running Meta Llama 3 and virtually any open‑source large language model locally, preserving privacy and offline availability.

Author: 万能的大雄 Reference: https://itnext.io/step-by-step-guide-to-running-latest-llm-model-meta-llama-3-on-apple-silicon-macs-m1-m2-or-m3-b9424ada6840