Introduction
My first foray into local AI on a phone was a rudimentary chat app in 2024, which was far from satisfactory. The app struggled with processing and lasted only a week before I deleted it. However, the landscape has changed significantly since then. Today, I've found five apps that have made running LLMs locally on a phone a practical and enjoyable experience. These apps integrate models like Qwen 3.5, Gemma 4, and LFM 2.5 into Siri, Shortcuts, and Control Center, allowing for quick and efficient interactions without the need for cloud services.
--## LM Studio
LM Studio, an app developed by Adrien Grondin, is a desktop app that runs models on computers. In 2026, LM Studio acquired LM Studio's official iPhone app, which runs models directly on the phone using Apple's MLX framework. No account is required, and the developer doesn’t collect any data. LM Studio supports a variety of models, including Qwen 3.5, LFM 2.5, Gemma, Llama, and Apple's own Foundation model. Qwen 3.5 can also handle images. Users can access Qwen 3.5 at the 2B size, and replies start almost immediately. When I tested it, I used Qwen 3.5 to compare NFL cities with the teams that call them home. The app was faster on the phone than on the Mac, and both versions produced thorough outlines, though they took different approaches.
--## LM Link
LM Link is a $4.99 app that pairs with LM Studio running on a home computer. It uses end-to-end encryption and allows users to run a bigger model on their phone without the heat or battery drain of running it locally. Chats are saved on the device, and users can flip between the phone and Mac versions for comparison. The app supports Qwen 3.5 9B, Phi 4, and Gemma, and newer Pro iPhones can also run Llama 3.1 8B. However, it only works with text and cannot handle photos.
--## Private LLM
Private LLM is a $4.99 app that integrates LLMs into Siri and the Shortcuts app. Users can use models as one step in an automation instead of opening the app and typing into a chat. Private LLM uses an MLC-LLM engine with OmniQuant quantization to keep answers sharper than the more basic compression used by other apps. The model list includes Qwen 3, Phi 4, and Gemma, and newer Pro iPhones can also run Llama 3.1 8B. Users can only dictate text, and Private LLM only works with text.
--## Conclusion
These five apps have made running LLMs locally on a phone a practical and enjoyable experience. They integrate models into Siri, Shortcuts, and Control Center, allowing for quick and efficient interactions without the need for cloud services. Whether you're looking to clean up voice notes, automate tasks, or just want a quick AI response, these apps are worth checking out.