Apple lets iPhones run AI with no internet, no bill

Something quietly big happened this week

At Apple's developer conference, the company announced that iPhones and Macs can now run a whole class of capable AI models — completely on the device, no internet required, no monthly bill.

Until now, most AI features in apps relied on sending your data to a server somewhere, waiting for a response, and paying per use. Apple has been chipping away at this with its own built-in model for a while, but this week they opened the door wider: developers can now bring in well-regarded open-source models like Qwen and Mistral and run them the same way, fully on Apple's own chip.

What does that mean in practice? An app could read a receipt, classify a photo, or summarise a document — entirely on your phone, without your data leaving it. For things like expense tracking, client intake forms, or internal tools, that's genuinely useful.

Apple also said the underlying framework will go open source later this summer, which means more eyes on it, more trust, and faster improvement.

This won't replace cloud AI for everything. But for a certain slice of tasks — the ones that are repetitive, private, or need to work offline — the economics just changed.

Words worth knowing

On-device — the AI runs on your phone or computer directly, not on a distant server. Like doing maths in your head instead of calling someone to do it for you.

Open-source model — an AI that anyone can inspect, download, and use. Think of it like a recipe that's been made public, versus a dish you can only order at one restaurant.

API key — a password that lets your app talk to an AI service. Usually tied to a billing account. On-device means you don't need one.

Foundation model — Apple's term for the AI built into every iPhone. It's already there, no download needed.