Apple has expanded its Foundation Models Framework (FMF) to include more sophisticated on-device and cloud-based large language models across macOS, iOS, iPadOS, and visionOS.
The updated framework improves model efficiency, reduces latency, and adds support for multimodal content generation alongside natural language understanding.
New developer tooling has been introduced to facilitate easier model deployment, performance tuning, and monitoring across all Apple hardware platforms.
The framework maintains a dual-architecture approach, utilizing lightweight on-device models for privacy and Private Cloud Compute (PCC) for larger, more complex tasks.
The rollout of these updated APIs is scheduled to occur throughout the 2026 fiscal year as developers integrate the new capabilities into their applications.
Apple’s strategy focuses on enabling high-performance AI integration while strictly adhering to the company's existing privacy safeguards.
That's the gist.
Dive into the full report for the data, charts, and sources behind these takeaways.
Read the full report