Apple and Meta in talks to bring Llama to iPhones

Apple just announced its massive partnership with OpenAI at WWDC, but Tim Cook is apparently not stopping there.

Meta has held talks with Apple about integrating its generative AI models directly into Apple Intelligence, according to a Sunday report from The Wall Street Journal. The two tech giants, notorious for years of public spats over user tracking and App Store rules, are now discussing how to share real estate on the iPhone.

A crowded waiting room for Siri

Meta is far from the only suitor knocking on Apple's door. The Journal reported that AI safety lab Anthropic has also been in discussions to plug its models into Apple Intelligence.

Meanwhile, a source told Reuters that AI search startup Perplexity is in similar talks. These discussions come alongside ongoing reports that Alphabet's Google is negotiating to bring Gemini to Apple devices as well.

Apple is also expected to pursue local AI partnerships in regional markets like China, where Microsoft-backed ChatGPT is unavailable.

None of these deals are finalized yet, and sources caution that talks could still fall apart. Meta and Anthropic declined to comment, while Apple did not immediately respond. While the exact financial terms remain unclear, the discussions involve AI partners selling premium subscription tiers directly through Apple Intelligence.

Why it matters

For everyday iPhone users, this is great news. Instead of being locked into OpenAI, you will likely get to choose your preferred digital brain directly inside iOS, whether that is Claude, Llama, Gemini, or ChatGPT. Getting to upgrade to premium tiers seamlessly through Apple's billing also takes the friction out of juggling separate AI subscriptions.

For the AI landscape, Apple is effectively turning Apple Intelligence into the App Store for foundational models. Building world-class frontier models is insanely expensive, but distribution is what turns models into businesses. By dangling access to over a billion high-value active devices, Apple holds all the leverage without having spent tens of billions training frontier LLMs itself.

We think this multi-vendor playbook is classic Apple pragmatism. By pitting OpenAI, Google, Meta, and Anthropic against one another, Apple avoids single-vendor lock-in while letting partners subsidize the brutal inference compute costs. The obvious hurdle is culture. Apple staked its entire AI brand on device-level privacy, an area where Meta has faced relentless scrutiny. Selling iPhone owners on a Meta-powered Siri feature might test just how far Apple's privacy halo can stretch.