How-to

How to download an Outlier model tier

Last updated 2026-06-18 · Outlier v1.11.469

What you need first for “how to download an outlier model tier”

Apple Silicon Mac, macOS 12 or later, the unified-memory minimum that the chosen tier requires (6 GB for Nano, 12 GB for Lite, 24 GB for Core / Code / Vision, 32 GB for Plus). Internet is required only for the one-time model download.

Steps

1. Open Settings > Models. Each tier shows its disk size and minimum RAM.
2. Click Download. The pull is HTTPS only; no auth token is needed for the public repos.
3. Wait for the green check. The app verifies the safetensors and tokenizer before marking the tier ready.
4. Switch tiers in chat. The active tier is shown in the bottom bar of the chat window.

What is the specific thing to know about this guide?

Each tier’s safetensors are pulled from a public Hugging Face repository under the Outlier-Ai namespace. The Plus tier is the exception: it pulls from mlx-community/Qwen3.5-397B-A17B-4bit at 209 GB.

What can go wrong with this guide?

• Hugging Face download interrupted partway leaves a half-checkpoint; the picker UI detects this on next launch and offers to resume.
• Slow networks make the Plus tier’s 209 GB pull effectively a multi-hour task.

How does this guide differ from the cloud equivalent?

Cloud-side model selection happens in a dropdown bound to a billing tier. Outlier’s model picker happens once, against on-disk weights you fetched yourself.

Each tier’s safetensors are pulled from a public Hugging Face repository under the Outlier-Ai namespace. The Plus tier is the exception: it pulls from mlx-community/Qwen3.5-397B-A17B-4bit at 209 GB.

What does this guide not claim about “how to download an outlier model tier”?

Each tier’s safetensors are pulled from a public Hugging Face repository under the Outlier-Ai namespace. The Plus tier is the exception: it pulls from mlx-community/Qwen3.5-397B-A17B-4bit at 209 GB.

This guide does not claim feature parity with cloud-side workflows for “how to download an outlier model tier”. Specifically, the product surface in v1.11.469 covers chat, file attachment, the local agent loop, project scoping, and the model picker. Cross-device sync, team workspaces, and shared session history are out of scope and are not on the v1.9 backlog either.

How does the model picker decide which tier is enabled?

The backend reads the Mac’s reported unified memory at startup and toggles each tier’s enabled flag based on a 15% slack against the tier’s min_ram_gb. The check lives in desktop_app/backend/server.py around line 1216 and is the reason Nano stays available on a 6 GB Mac while Core appears greyed out on the same hardware.

Where does this guide fit in the rest of the lineup?

The Plus tier is the outlier on size: 209 GB versus the next-largest Vision at 19 GB. Most users start with Nano or Lite, then graduate to Core or Code once they have a use case that exercises the heavier weights.

One unique number

4 steps, zero network requests after the model is downloaded. Tier sizes range from 2.37 GB (Nano) to 209 GB (Plus); the Quick tier is 15.61 GB and Core is 15.13 GB.

Download Outlier for Mac

Requires Apple Silicon (M1, M2, M3, or M4) — Intel Macs are not supported. macOS 12+.