PrivateLM

A local AI chat client that runs GGUF models offline on Android and iOS, supports image conversations, and connects to cloud APIs.

Alternative to:

ChatGPT

PrivateLM is an AI chat app for people who want to run models on their phone or tablet and choose when to use a cloud provider. It supports local inference on Android and iOS; the browser version requires cloud APIs.

Offline model chat: Run downloaded GGUF models on your device through llama.cpp, with Vulkan GPU acceleration on Android and Metal on iOS.
Image conversations: Include images alongside text, with local vision support through Qwen2-VL and vision capabilities through cloud providers.
Cloud provider choice: Switch to OpenAI, Anthropic, Google Gemini, or Kimi when you want to use an external service rather than a model on your device.
Local chat storage: Keep conversations and settings on the device; choosing cloud mode sends requests to the selected provider, and API keys go only to that provider's endpoint.

PrivateLM is open source under the MIT license. Local model chat works without internet after you download a model, while cloud chat needs a connection and provider API keys. Android local inference targets ARM64 devices, and the app uses available RAM to recommend context and response limits.

The iOS app is distributed as a sideloadable IPA. It runs on iPhones and iPads, though iPad is the recommended target because local models need substantial RAM.