Apple offers free on-device AI to developers, eliminating API costs

Apple has announced that developers can now access its on-device AI large language model for free through the Foundation Model Framework across all its operating systems. This removes both the financial barriers and privacy concerns that have previously made incorporating AI features into apps a complex business decision, potentially transforming how developers approach app creation.

Why this matters: Unlike existing AI services that charge per API call, Apple’s approach eliminates transactional costs and keeps data processing entirely on-device, removing the two biggest friction points that have limited AI adoption among smaller developers.

The big picture: This mirrors the revolutionary impact of the original App Store launch, which removed distribution barriers and enabled anyone to create profitable apps without massive upfront costs.

• Before the App Store, developers faced tens of thousands in packaging costs, complex distribution channels, and kept only 30-40% of revenue.
• The App Store eliminated production costs and gave developers 70% of revenue, enabling the creation of millions of apps.
• Similarly, AI integration has required careful return-on-investment calculations due to ongoing API costs and privacy concerns.

How the current AI landscape works: Most developers face four challenging business models when adding AI features to their apps.

• Charge users a monthly subscription specifically for AI features.
• Bury AI costs in existing monthly fees, reducing profit margins.
• Implement per-usage credit systems that create user friction.
• Absorb API costs entirely, risking unprofitable “whale” users who consume excessive resources.

What Apple is offering: The Foundation Model Framework provides on-device AI processing at no charge to developers.

• Developers can add AI prompts as easily as adding dropdown menus or calendars to their apps.
• Privacy is preserved since all processing happens locally on users’ devices.
• Features include smart tagging, image recognition, thumbnail generation, and enhanced game dialogue.

Real-world impact: Paul Mayne, head of Day One at Automattic, a WordPress company, highlighted the privacy benefits for sensitive applications.

• “The Foundation Model framework has helped us rethink what’s possible with journaling,” Mayne said.
• “Now we can bring intelligence and privacy together in ways that deeply respect our users.”

Additional developer tools: Apple also announced AI integration in Xcode 26, its development environment.

• Developers can use any large language model they prefer, including ChatGPT or Anthropic models.
• Local model execution prevents code from being sent to cloud-based AI services.
• This addresses security concerns for developers working under non-disclosure agreements or code-sharing restrictions.

The reality check: Apple Intelligence as a consumer product remains underwhelming compared to competitors like OpenAI, Google, and Microsoft.

• Apple’s AI models are currently described as “fairly meh” but are expected to improve over time.
• The company focused more on user interface improvements than breakthrough AI capabilities at its recent developer conference.