Private AI, from pocket to desktop.

See the app. See the workflow.

A three-minute ModelDeck walkthrough, from the native interface to AI workflows. The demo shows the product experience; local and connected capabilities are explained on the Apple apps page.

ModelDeck: a native AI interface

ModelDeck is a SwiftUI iPhone/iPad application for chat and private model management. It brings on-device models and remote model services into one application, while distinguishing where a workload runs.

On-device MLX on iPhone

The implementation integrates MLX language and vision models, model downloads and imports, and local speech workflows. Memory-aware admission checks, a bounded buffer cache, idle unloading and background handling are part of the model lifecycle.

I’ve tested text and vision generation and the local speech pipeline on a physical iPhone. Model choice depends on the device’s memory and the workload.

iPad and MacBook applications

I’ve also built apps for iPad and MacBook, and I run MLX models on my MacBook. My work covers the native interface, model runtime and the complete user workflow.

Local inference and connected infrastructure

On-device inference runs a compatible model on the Apple device. Connected workflows call a model server on another machine. ModelDeck supports private infrastructure connections and includes native image and video forms backed by remote generation services; those media services should not be confused with on-phone generation.

What this enables for a client

• A native interface designed around a useful AI task.
• On-device inference where model size and device resources fit.
• Private-server inference when the workload needs more compute.
• Explicit data routing, model lifecycle controls and device-specific evaluation.