Choosing a model for your Mac

Osaurus works with dozens of models. You don't need to understand them all. This guide gives you a quick way to decide, shows how much your Mac can handle, and explains how to switch.

The short answer: start with Raptor 0.7. It's Osaurus's own model, built for agents and workflows, runs on every supported Mac, and needs about 3.7 GB of space. If you chose Keep it private, on this Mac during setup, you already have it.

Decide in ten seconds

Ask yourself what matters most.

• Privacy and working offline: use a local model. Raptor 0.7 is the place to start.
• The strongest answers, with nothing to download: use Osaurus Cloud. You pay with credits, and new users get free credit to start.
• You already pay for ChatGPT or Claude: sign in with that account in Settings → Providers, or use Claude Code. No new subscription needed.

Will it fit on my Mac?

Local models use your Mac's memory (RAM) while they run. The more memory your Mac has, the bigger the model it can run comfortably. To check yours, open the Apple menu and choose About This Mac. The Memory line is the number you want.

A rough guide:

• 8 GB Mac: stick to small models like Raptor 0.7 or LFM3-VL 2B.
• 16 GB Mac: small and mid-size models up to about 8 GB run comfortably.
• 32 GB or more: larger models of 20 GB and up become an option.

You don't have to do this math yourself. In Settings → Local Models, every model shows a Fits This Mac label.

Tip: Model names hint at size. "4B" means 4 billion parameters. A name like "35B-A3B" means only a small part of the model works at a time, so it's lighter on memory than its size suggests.

Download a different local model

1. Open Settings → Local Models. Click the gear at the bottom of the narrow strip on the left, or press ⌘ ,.
2. Browse or search the list. Each row says whether it Fits This Mac, and Top Picks are marked with a star.
3. Click Download. You can pause and resume.
4. Pick it in a chat. Click the model name in the message box and choose your new model.

Not sure which one? Just ask in a chat:

Osaurus can recommend a model for your Mac and download it for you, after you approve.

Popular picks

Want a recommendation based on your exact Mac? Try the model picker tool.

Switch models any time

Switching takes one click and you lose nothing. Your agents, memory, and tools carry over.

1. Click the model name in the chat's message box.
2. Pick a provider on the left: Local, Osaurus Cloud, or one you've connected.
3. Pick a model on the right.

Some models show a third column with extra options, like Thinking, which makes the model work through a problem step by step before answering.

Quick answers

Is a bigger model always better? Not for everyday tasks. A smaller model replies faster and leaves room for your other apps. Go bigger when answers feel shallow.

Does a cloud model see my memory and files? It sees whatever is sent with your message, which can include memory, document excerpts, and tool results. A local model keeps all of that on your Mac.

What about Apple Intelligence? On macOS 26 or later, pick Foundation in the model picker. Nothing to download. It's small, so it's great for quick questions but can't build workflows.

Where are models stored? In a folder called MLXModels in your home folder. Delete one from Settings → Local Models → On This Mac.

Now put it to work: start with the Morning Brief, a one-click summary of your day.