Bring an LLM provider to the Foundation Models framework

A Podkey summary of Bring an LLM provider to the Foundation Models framework, from WWDC 2026.Apple’s doing something pretty practical with AI here: instead of making developers pick one model stack and marry it forever, it’s building a common way to talk to lots of models. The big idea is one protocol, one app-facing API, and then a mix of options underneath it, from fully local models on your device to Apple’s private cloud to outside providers like Anthropic, Google, and community models. And once you get past the framework jargon, the story is really about flexibility, privacy, and making Swift apps a lot less painful to upgrade as the model landscape keeps changing every five minutes.One protocol for every modelApple’s local model gets betterWhen local isn’t enoughMore model choices, not fewerHow the shared API actually helpsSpeed, caching, and avoiding first-request awkwardnessStreaming, errors, and richer responsesTools, agents, and real app behaviorSecurity and trustThis podcast was created with Podkey. Make your own at https://podkey.fm