Locally AI

Locally AI is a native Apple app that lets you run powerful open-source language and vision models entirely on-device — no internet, no login, no cloud. It supports a wide range of models including Meta Llama, Google Gemma, Qwen, DeepSeek, and more, all optimized for Apple Silicon via the MLX framework. Your data never leaves your device, making it one of the most privacy-preserving AI assistants available for iPhone, iPad, and Mac.

• Offline Operation — Download a model once and use it indefinitely without any internet connection.
• Complete Privacy — All inference runs locally; no data is sent to external servers or collected.
• Apple Silicon Optimization — Built on Apple's MLX framework to maximize speed and efficiency on A-series and M-series chips.
• Local Voice Mode — Hold real-time voice conversations with the AI entirely on-device, no cloud processing required.
• Language and Vision Models — Supports both text and image processing tasks using state-of-the-art multimodal models like Qwen 2 VL.
• Siri Integration — Trigger on-device AI conversations directly via Siri with "Hey, Locally AI."
• Customizable System Prompt — Tailor the AI's personality and behavior to your specific use case.
• Apple Shortcuts Integration — Automate tasks and create custom AI-powered workflows using the Shortcuts app.
• Controls Integration — Access the app from Control Center, Lock Screen, or the Action Button for instant use.
• Wide Model Support — Runs Llama 3.1/3.2, Gemma 2/3/3n, Qwen 2/2.5/3, DeepSeek R1, SmolLM, IBM Granite, Cogito, and Liquid LFM.
• iPad & Mac Support — Run larger models on bigger devices for more advanced tasks, rivaling cloud-based GPT-4 class models.