This post is from the WWDC26 watchOS Q&A.
Hi! I've seen during the Keynote that watchOS will support Siri AI (when your iPhone is nearby), which sounds really exciting.
Out of curiousity, I also noticed that both Core AI, and now the Foundation Models framework, are available on watchOS 27. This sounds amazing - bringing intelligence features to the watch could enable some great tailored experiences on the go.
Here are some of my questions:
- Do these require a nearby iPhone?
- Do they work offline (on-device)?
- Lastly, are there any device limitations?
- Also, if there are any additional details or fun facts about using Foundation Models or Core AI on watchOS (or perhaps iOS, if making use of the companion app to run those models on the iPhone), I’d love to hear them!
We're also really excited about these. For the Foundation Models framework:
The Foundation Models framework on watchOS is great for glanceable summaries, smart reply suggestions, and simplifying dense text.
- It does not require a nearby iPhone.
- It does not work offline. It requires a network connection.
- It works on all watchOS 27 devices.
- You can use the Private Cloud Compute language model (with a managed entitlement) or a third-party language model.
- See these WWDC sessions for a lot more great information: Build with the new Apple Foundation Model in Private Cloud Compute and What's new in the Foundation Models framework
For the CoreAI framework:
- This does not require a nearby iPhone. This is for on-device models.
- It does work offline because the models are running on device.
- I believe that this works with all watchOS 27 devices.
- Apple Watch is a small device with a limited battery. So this isn't the place to run larger models. For things like random forests/tree ensembles, categorical feature engineering, and pipelines, continue to use CoreML on watchOS. But if you want to use iterative inference loops on watchOS, you might want to explore the CoreAI framework on watchOS.
