Show HN: Gemma 4 26B A4B running on an iPhone 17 Pro via model paging (noemaai.com)

A new application called Gemma has been launched, allowing users to run advanced AI models directly on their iPhones, iPads, Macs, and Vision Pro devices. This app introduces a groundbreaking privacy model by enabling local, on-device processing, ensuring that user data and conversations remain private. With features like off-grid mode that blocks external network access, Gemma empowers users to manage their AI tasks without reliance on the cloud, fostering greater control over AI interactions. Significantly, Gemma allows users to leverage various model types, including GGUF, MLX, Core ML, and Apple's Foundation Models, all within a single native environment. The introduction of Noema 2B, a compact 2-billion-parameter model optimized for efficient reasoning and task execution on personal devices, underscores a shift towards more powerful yet accessible AI capabilities. This innovation not only elevates personal productivity by allowing seamless access to local files and data for thorough analysis but also paves the way for a more private, user-centric approach to AI, challenging traditional cloud-dependent models in the AI/ML community.