iPhone 17 Pro ran a 400B parameter large language model (LLM) during a WWDC 2025 developer session, per a Hacker News post. The demo, shared via Twitter by user @anemll, showed the device handling complex AI tasks locally without cloud dependency. Comments on Hacker News speculate this could redefine on-device AI performance, though Apple hasn’t confirmed details.

The demo aligns with Apple’s history of optimizing hardware for software breakthroughs. By showcasing a 400B LLM—roughly 10x larger than current iPhone capabilities—Apple may be testing edge computing limits ahead of 2025 product cycles. Developers on Hacker News noted the demo’s potential for real-time language processing, though specifics about model efficiency or training data remain unclear.

This move underscores Apple’s push to compete with Google and Microsoft in AI-driven developer tools. If accurate, the 400B LLM could enable breakthroughs in localized AI applications, from coding assistants to privacy-focused data handling. However, skeptics argue the demo might rely on proprietary optimizations rather than raw computational power.

Key takeaway: The iPhone 17 Pro’s rumored AI prowess signals Apple’s ambition to lead in on-device machine learning. While details are sparse, the demo’s implications for AI development workflows and consumer hardware innovation warrant close attention.