I can confirm the behavior and want to add the developer perspective. I build an app that uses Apple’s on-device Foundation Models (with a Core ML fallback). When the device UI is set to Catalan, Apple Intelligence is disabled, so the Foundation Models APIs report unavailable. This blocks my app from handling prompts in English or Spanish even when the user explicitly selects those languages in-app. Developer impact • I must ship separate code paths and degrade features based on UI language, not user intent. • Users who prefer a Catalan UI lose on-device, privacy-preserving LLM features, so I am forced to route to cloud or third-party SDKs. • QA complexity rises because availability hinges on language toggles and reboots, as the bug report here illustrates. Repro (app context) 1. Set iPhone language to Catalan. 2. Call Foundation Models API from the app. availability reports not available. 3. Switch device language to Spanish. Availability often remains “unavailable” until reboot. 4. After reboot, availability is restored. Switching back to Catalan may keep it enabled until the first Catalan query, then it disables again. Expected • I can keep the device UI in Catalan and still request English or Spanish understanding/generation for my app. • Availability should not require a reboot when moving to a supported language. Requests 1. Decouple Apple Intelligence availability from the device UI language. Let users enable the feature if they choose any supported model language. 2. Provide a per-app model language override so apps can request EN/ES while the UI remains CA. 3. Make availability reasons precise and actionable when blocked due to language policy. 4. Fix the state inconsistency after language switches to remove the reboot requirement. Happy to share diagnostics, sample code, and a minimal project if that helps.