Apple's On-Device Intelligence Adds Third-Party Model Support
What Happened
The update, previewed at Apple's spring developer briefing, introduces a Private Model API allowing vetted third-party models to run in the Secure Neural Engine. Apple confirmed Mistral Small 3 and Microsoft's Phi-5 as launch partners, with Gemini Nano and Claude Haiku in talks. All models remain sandboxed from network access by default.
My Take
Apple finally conceding it can't win on pure model quality is a pragmatic pivot, but the bigger story is that on-device inference is about to become a real distribution channel. For enterprise software vendors, this means you can ship a model alongside your iOS app that runs entirely privately — huge for healthcare, legal, and finance use cases where cloud inference is a compliance headache. Expect a flurry of B2B apps repositioning around "on-device AI" in the next six months.
Read Original Source