Personal LLM

Personal LLM runs AI chat, photo analysis and document questions on your phone, with no account or cloud inference. It's for people who want a private ChatGPT alternative on iOS or Android, including when they don't have a connection. Once you've downloaded a model and any image support files, those features work offline.

The catalog includes Qwen, Gemma, GLM and Ministral models. You can also add GGUF models by URL, with an optional vision projector for image input. The app uses llama.cpp, with Metal acceleration on iPhone, OpenCL on supported Snapdragon Adreno phones and a CPU fallback. Model cards check your phone's RAM: the smallest model can run with 2 GB, while larger choices need 8 GB or more.

Photos can include menus, forms or screenshots. For PDF, text and Markdown documents, the app searches the text locally and identifies the passages used in its answer. Supported reasoning models can show their working in a collapsible panel. You can switch models within a conversation, adjust response settings and give each chat its own system prompt. On-device text-to-speech reads answers aloud, and file exports let you back up chats and settings.

Model downloads, advertising and purchase checks use the internet; Google AdMob may collect device information. Chats and attached photos stay on the phone unless you report an answer to support, which sends that answer and your note. The app excludes its files from iCloud and Google backups and doesn't provide cloud sync.