Apple announced a new AI architecture built on Google Gemini models. The architecture runs both on-device and on servers, supporting multimodal capabilities like image understanding and generation.
Apple announced a major overhaul of its Apple Intelligence platform on June 8, 2026, revealing a new AI architecture built on foundation models co-developed with Google using technologies behind the Gemini family. The architecture runs both on-device and on servers via Apple's Private Cloud Compute infrastructure.
Apple chose to collaborate with Google to overcome limitations of its previous AI platform. The new models support multimodal capabilities including realistic image creation, advanced photo editing, and visual question answering. Some devices will receive additional features like speech generation, improved dictation accuracy, and stronger natural language understanding. A new system orchestrator coordinates features across platforms, tailoring responses based on the active app and user task.
Apple framed its approach as a contrast to competitors it characterized as racing forward without regard for users, emphasizing on-device processing and Private Cloud Compute to protect user data. This highlights Apple's strategy to differentiate itself in the AI race by prioritizing privacy.
Apple unveiled five AI models based on Google Gemini in iOS 27, with the highest-performance Cloud Pro model noted to be a Gemini frontier-level model running on NVIDIA GPUs in Google Cloud. However, many commenters criticized the frequent hallucinations and overly limited performance of the public Gemini version, expressing concerns that Apple's reliance on it could lead to quality issues. In particular, it was pointed out that AI mode in Google Search has provided incorrect answers, worsening public perception.