How to run a local LLM on your Mac, without the terminal
Local models used to mean installing a runtime, hunting for the right file and typing commands. This guide gets a private model answering on your Mac in a few minutes, with clicks only.
Before you start
You need an Apple Silicon Mac with macOS 11 or later, and some free disk space: from about 400 MB for the smallest models to about 19 GB for the largest. Pandore also runs local models on Windows 10 and 11 with the same steps.
Nothing else is required. You do not need Homebrew, Python, Ollama or a separate llama.cpp install.
Step 1: Open the local model library
- Install Pandore and open it.
- Go to Settings › AI and choose the local mode.
- Follow the link to the Local models library.
The library shows your Mac's processor and memory at the top. Use the recommended filter to see only the models that fit your memory.
Step 2: Pick a model
If you are not sure, start small. A model with 1 to 4 billion parameters answers quickly on any Apple Silicon Mac and handles everyday text work well.
| If you want | Try |
|---|---|
| Fast everyday answers | Qwen 2.5 1.5B or Gemma 3 1B |
| Better writing and analysis | Gemma 3 4B or Qwen 3 4B |
| French and other European languages | Ministral 3 8B |
| Step-by-step reasoning | DeepSeek R1 Distill 7B or a Ministral 3 Reasoning model |
| Help with code | Qwen 2.5 Coder 7B |
Each card shows the model's size, the memory it needs and what it is best for.
Step 3: Download it
Click Download on the model you chose. Pandore downloads a single file and, the first time, installs the runtime it needs. When the download ends, the model is ready to use.
Step 4: Ask your first question
Open the chat, or a project's conversation, and ask something. The model runs on your Mac: your question, your documents and the answer never leave it.
Pandore loads the model in the background and keeps it ready while the app is open, so you do not wait for it to start every time.
Step 5: Let Pandore choose local or remote
Local models are best for focused text work. For tasks that need tools, such as working inside a project folder or running a Mission, you need a remote agent like Claude Code.
In Settings › AI, the Both mode does this for you with automatic routing: Pandore keeps simple or sensitive requests local and uses your remote agent when a task needs tools or more capacity. The AI usage dashboard then shows how much work stayed on your Mac.
