Apple used its WWDC 2026 keynote to clarify that the new Foundation Models powering Siri and Apple Intelligence contain zero Google Gemini code. Despite earlier announcements about partnering with Google, the company emphasized these are entirely custom-built models trained on proprietary data. The clarification comes after months of speculation that Apple had essentially given up and was just rebadging Gemini technology.

The Foundation Models lineup includes AFM Core, AFM Core Advanced (which features native multi-modal capabilities), AFM Cloud for server-side processing, AFM Cloud Image for generation tasks, and AFM Cloud Pro for demanding agentic tools. All models are optimized specifically for Apple Silicon. While Google's infrastructure and NVIDIA GPUs support AFM Cloud Pro, Apple maintains the Private Cloud Compute certification and allows third-party verification of its data handling practices.

This approach lets Apple leverage external technology for training while maintaining its privacy-focused brand promise. The company acknowledged early missteps with Apple Intelligence, admitting it overpromised on features that couldn't deliver due to AI limitations. Going forward, the iPhone 17 Pro Max will showcase the most powerful iterations of these models.

Apple's strategy demonstrates how tech giants can collaborate on AI development without compromising their core product identity. By using distillation techniques rather than direct integration, Apple preserves its software sovereignty while accelerating development timelines.