Frequently asked questions

1. Is it really free?
2. Does it work offline?
3. How good is it compared to ChatGPT or Claude?
4. Which Macs does it run on?
5. Will it slow down my Mac?
6. Where are my chats stored?
7. Is Windows or Linux supported?
8. Can I use other AI models?
9. The first reply is slow
10. Voice mode keeps interrupting itself
11. How do I free up space?
12. How do I update?
13. How do I report a problem?

Is it really free?

Yes. The app is open source (Apache 2.0) and the AI models are free to download. There’s no subscription and no usage limit: it runs on hardware you already own.

Does it work offline?

Yes, once installed. Chatting, voice, files and video creation all work without internet. Only web search and opening web pages need a connection.

How good is it compared to ChatGPT or Claude?

The main model (Qwen 3.5, 35 billion parameters) is very capable for everyday questions, writing, summarizing, coding help and understanding images. The largest cloud models are still stronger at the hardest reasoning and specialized knowledge. You trade some of that for privacy, no cost and offline use.

Which Macs does it run on?

Macs with Apple silicon (M1 or newer), 32 GB of memory or more, and macOS 15 or newer. With 64 GB or more you can choose the default model, Qwen 3.5, and add video creation. With 32–63 GB the installer sets it up with Gemma 4, a lighter model that still handles chat, photos, documents and voice; 48 GB Macs can add LTX-2.3 video. It’s developed and tested on a MacBook Pro with an M5 Max and 64 GB. See the full table, and help us test other sizes.

Will it slow down my Mac?

While the app is open, the AI model uses about 20 GB of memory (about 8 GB with Gemma) and the graphics chip works while it’s answering. When you quit the app, all of that is released. Creating a video uses the chip fully for a minute or more.

Where are my chats stored?

In ~/.gorunrun/local/data on your Mac. Incognito chats and their files are kept only until you quit the app, then deleted.

Is Windows or Linux supported?

Not yet. It relies on Apple’s MLX framework for speed. Support for other platforms is planned; follow the project on GitHub.

Can I use other AI models?

Yes. Developers can add or swap models in one configuration file; see Extending it.

The first reply is slow

The first reply after opening the app waits for the model to load (up to a minute). After that, replies start in well under a second.

Voice mode keeps interrupting itself

The microphone is hearing the speakers. Use headphones, lower the volume, or lower the sensitivity in Settings → Voice.

How do I free up space?

Open Settings → Clear clutter. There are two separate groups, each for everything or a date range: Generated by AI (videos, images and files it made, mini apps and diagrams, temporary files) and Your uploads (photos and screenshots, videos, documents, voice messages). It shows how much each will free before anything is removed. The text of your chats is always kept, and files you added to a Project stay.

How do I update?

Open Settings → Updates and click Check for updates. If there’s a new version, it shows what’s new and the one command to run in Terminal: the same one you installed with. Updating keeps your chats, settings and downloaded models. The app never checks by itself.

How do I report a problem?

Open an issue on GitHub. Include your Mac model and memory, and the last lines of the log (Help → Show Logs in the app).