On this page
Every small Qwen 3.5, from 0.8B to 9B, runs on an Apple silicon Mac. Off Grid AI Desktop lists all four sizes and runs them for free and offline. On a 16 GB Mac, run Qwen 3.5 9B. On an 8 GB Mac, run Qwen 3.5 4B, or Bonsai 2 9B, a 3 GB build of the 9B.
All of them read images, think before they answer, and keep working with Wi-Fi off.
Download Off Grid AI Desktop (free, no account)
Which Qwen 3.5 fits your Mac
Off Grid AI Desktop checks every download against your Mac’s memory. A model up to 65% of your memory Runs well. Up to 82% it’s Tight, which means it loads but leaves little room for long chats. Sizes below include each model’s vision file.
| Model | Size | 8 GB Mac | 16 GB Mac or more | What it’s for |
|---|---|---|---|---|
| Qwen 3.5 0.8B | 0.7 GB | Runs well | Runs well | Tiny and very fast. Short rewrites and quick lookups |
| Qwen 3.5 2B | 2.0 GB | Runs well | Runs well | Fast chat with thinking and long context |
| Qwen 3.5 4B | 3.4 GB | Runs well | Runs well | The strong small all-rounder |
| Qwen 3.5 9B | 6.6 GB | Tight | Runs well | Better reasoning, needs more memory |
| Bonsai 2 9B | 3.0 GB | Runs well | Runs well | Qwen 3.5 9B packed small for 8 GB Macs |
The sizes are the 4-bit files in the app’s catalog, from Unsloth’s Qwen 3.5 files. The verdicts are the app’s own rule applied to those sizes. The “What it’s for” column follows the app’s own description of each model.
On a 16 GB Mac you can also go up to Qwen 3.8 27B as Bonsai 2 27B. See How to Run Qwen 3.8 27B on Your Mac.
What Qwen 3.5 is
Qwen 3.5 is Alibaba’s February 2026 family, under the Apache 2.0 licence. Every size reads images as well as text. They think before they answer by default, handle 262,144 tokens of context natively, and Alibaba says they cover 201 languages and dialects (model card).
On a Mac, Off Grid AI Desktop runs them through llama.cpp on Metal, so your Mac’s GPU does the work.
Bonsai 2 9B is Prism ML’s build of Qwen 3.5 9B with every weight stored as -1, 0 or +1. That takes the language model from about 17.9 GB to 2.4 GB. The app ships Prism’s engine for it on the Mac, so there’s nothing to build.
How to set it up
- Download Off Grid AI Desktop, install it and open it.
- Open Models > Text.
- Download the size for your Mac:
- 16 GB or more: Qwen 3.5 9B.
- 8 GB: Qwen 3.5 4B, or Bonsai 2 9B for the 9B’s reasoning.
- Wait for the download. It’s 0.7 to 6.6 GB.
- Open a new chat, select the model and ask something you can check.
The 4B and 9B cards show Faster replies. They come with a small draft helper (DFlash) that guesses the next words, so answers arrive sooner when the model fits in GPU memory.
Try a task that fits a small model:
Rewrite this message so it sounds friendly but firm. Keep it under 60 words.
Then paste a message you’ve been putting off. To ask about a picture, attach an image to the chat. Turn off Wi-Fi and ask again. It keeps answering, because nothing leaves your Mac.
Which size to keep
- Use the biggest one that says Runs well. On 16 GB that’s the 9B, which the app’s own quality ranking puts well above the 4B.
- Keep the 0.8B or 2B for quick jobs. They answer almost instantly, which suits rewrites, short summaries and tagging.
- On 8 GB, try Bonsai 2 9B against the 4B. Ask both the same question and keep the one whose answers you trust.
If something goes wrong
| Symptom | Check | Fix |
|---|---|---|
| The 9B card says Tight | Your Mac has 8 GB | Use Qwen 3.5 4B or Bonsai 2 9B |
| Replies are slow | Other apps using memory | Close heavy apps, or drop a size |
| Long chats get slow | Context window in the model’s settings | Lower it, or start a new chat |
| Image questions fail | The vision file | Check it finished downloading |
| You searched for Qwen 3.5 27B | The card says Newer: Qwen 3.8 27B | Use Qwen 3.8 27B, which does the same job better |
FAQ
Which Qwen 3.5 should I run on a 16 GB Mac? Qwen 3.5 9B. It’s a 6.6 GB download with its vision file, and Off Grid AI Desktop marks it Runs well on 16 GB. Every smaller Qwen 3.5 fits too.
Can an 8 GB Mac run Qwen 3.5? Yes. The 0.8B, 2B and 4B run well. The standard 9B is Tight on 8 GB, so use Bonsai 2 9B instead, a 3 GB build of Qwen 3.5 9B that the app lists for 8 GB.
Does Qwen 3.5 work offline on a Mac? Yes. After the download, it runs on your Mac with Wi-Fi off. Your chats and files stay on your disk, and there’s no account.
Can Qwen 3.5 read images? Yes. Every Qwen 3.5 size reads images, and Off Grid AI Desktop downloads its small vision file with the model. Attach an image to the chat.
What about Qwen 3.5 27B and 35B-A3B? Newer models do those jobs better. Off Grid AI Desktop points Qwen 3.5 27B to Qwen 3.8 27B, and Qwen 3.5 35B-A3B to Qwen 3.6 35B-A3B.
Related guides
- How to Run Qwen 3.8 27B on Your Mac in 2026
- How to Run Qwen 3.6 35B-A3B on Your Mac in 2026
- How to Run Qwen 3.5 on Your iPhone in 2026
- How to Run Qwen Locally on Your Computer in 2026
Download Off Grid AI Desktop and run Qwen 3.5 on your Mac today.