Understanding Foundation Models
What Apple's on-device AI is, and how Bit Buddy uses it to analyze your devices.
Starting with iOS 26, Apple introduced Foundation Models: a set of AI models that run directly on your device's hardware. Bit Buddy takes advantage of these models to provide storage analysis and device recommendations without sending any of your data to a server.
If you've used AI tools like ChatGPT or Gemini before, the concept is similar, but with one big difference: everything happens on your iPhone or iPad. No internet connection needed, no data leaves the device, and no third party ever sees your information.
How It Works in Bit Buddy
When you open an AI analysis feature in Bit Buddy, the app sends a prompt to Apple's on-device model along with relevant information about your devices. The model processes everything locally and returns a response, usually within a few seconds.
Bit Buddy uses Foundation Models in a few places:
- AI Insight Cards: These appear on device detail pages and give you a quick analysis of that specific device, like storage patterns or maintenance suggestions.
- AI Tips: Personalized recommendations based on your entire device collection, covering things like backup habits, storage optimization, and device health.
- Ask Buddy: A conversational interface where you can ask questions about your devices in plain English.
On-Device vs. Cloud AI
Bit Buddy supports two AI modes: on-device (Foundation Models) and cloud (OpenRouter). Here's how they compare:
On-device (Foundation Models):
- Requires iOS 26 or later
- Runs entirely on your device's processor
- Works offline
- Your data never leaves the device
- No usage limits or API costs
Cloud AI (OpenRouter):
- Works on any supported iOS version
- Requires an internet connection
- Data is anonymized before being sent (device names, dates, notes, and locations are stripped or generalized)
- Opt-in only
If your device supports Foundation Models, Bit Buddy will default to on-device processing. You can switch between the two in Settings under the AI section.
Privacy
This is worth emphasizing: when using Foundation Models, absolutely nothing leaves your device. Apple's models run in a sandboxed environment on the device itself. Apple can't see your queries. Bit Buddy can't see your queries. The data stays in memory during processing and isn't stored or transmitted anywhere.
For users who care about keeping their device data private (which, if you're reading this, probably includes you), Foundation Models are the ideal option.
Limitations
Foundation Models are impressive, but they aren't perfect. A few things to keep in mind:
- The models are smaller than cloud-based alternatives, so responses can occasionally be less detailed.
- Processing speed depends on your device's hardware. Newer iPhones and iPads will be faster.
- Foundation Models require iOS 26, so older devices that can't update won't have access.
- The model can sometimes misinterpret context, especially with large or complex device collections.
That said, for most users, the on-device experience is more than capable for day-to-day device analysis and recommendations.