How to use Apple Foundation models in BTT?

Configuration Help & General Discussion
1

How do I configure BTT to use an Apple Foundation model?

I'm not seeing it listed anywhere in the AI Config window. Where would I look?

  1. I'm on macOS 27.0.1 and BTT 6.870, with an M4 Pro
  2. I have Siri turned off because I don't want to bother with it in most apps.
  3. But in several other apps I've still been able to access the Apple Foundation Advanced model in other dedicated apps like https://locallyai.app (since it's still on the system anyway).

I don't know how useful the model in BTT, but I thought I'd give it a try.

2

In my tests the local Apple Foundation Models are barely usable, but you can configure it in BTT's AI setup wizard

image
image2324×1448 188 KB

I'd love to be able to provide access to the private cloud compute models, but Apple doesn't allow this for apps distributed outside the app store.

3

I figured they may not be very usable, but I thought I'd try.

For the Local Model Configuration, I don't get the option for Apple On-Device:

Screenshot 2026-09-29 at 8.37.32 AM
Screenshot 2026-09-29 at 8.37.32 AM1824×1448 175 KB

I'm guessing it might be because Siri is turned off?

I originally thought that having Siri turned off would disable AI everywhere on the system (even if you couldn't reclaim the storage space). But it still worked with apps like LocallyAI. And I found lots of other places where I had to actually disable things anyway (like Writing Tools, Mail and Messages summaries, etc.)

Since it's still there, figured I might as well try it out.

4

Ah I think you selected the BetterTouchTool Configuration wizard, for that specific type of assistant the model is disabled completely because it 100% produced only non-working solutions :see_no_evil_monkey:

Standard AI Assistants should show it:

image
image1920×1383 212 KB

5

Thanks for the explanation!

I do see it under the standard AI Assistants.

6

Related:

I do have LM Studio installed, but I haven't wanted to keep it running all the time (so as not to use up lots of RAM, for example). I have a couple models downloaded to an external SSD, and then start it when I actually am doing something I think might benefit from it. Then I shut it down.

Would it be possible to configure BTT to only use it when I wanted to do things like create plugins?

7

BTT will not do anything with the connection unless you use it. You can always shut down the LM Studio (I think there are now faster tools for specific models, I have experimented with oMLX and Qwen 3.8 27b, which has been pretty good)

I will upgrade to macOS 27.2 beta later today to see whether the latest Apple foundation models got any better (apparently they are now available again in europe)

8

Thanks!

That's helpful to know.

And thanks for the tip about oMLX. It looks interesting. I actually only finally decided to try out local AI models last week, when I figured out my MacBook could probably handle Qwen 3.8 27b.

LM Studio seemed the easiest to set up, but I haven't been overly thrilled with it, but also not terribly disappointed, since I'm relatively new to all of it.

Overall I've been slow to try out AI, but local models seem more appealing.

9

unfortunately on most Macs these models are still too slow to be really usable. (Some might disagree but I find them just too slow to use)
With the latest Mac Studio Ultra this changed but that one is also super expensive :slight_smile:

10

I wouldn't mind things being a bit faster, but since I haven't had much experience with all the normal (non-local) AI models out there, my expectations are probably not as high. :slight_smile:

11

Is there a way to configure the max conversation time another value now?

12

you were able to get Writing Tools to show up in Golden Gate Beta? Damn for me it seems they replaced it with Ask Siri and while the spotlight is nicer and such i like the wriiting prompts menu instead of "speaking" or typing into sri what i what to do with text and sometime it offers suggustions list below but not frequently