Private LLM

Private LLM is an AI chat app for iPhone, iPad, and Mac that runs large language models entirely on the device, with no cloud connection, account, or internet access required after the initial model download. It supports more than 140 open-source models, including Llama 3.3 70B, Phi-4, Google Gemma 3, Qwen3, and DeepSeek R1 Distill, using OmniQuant and GPTQ quantization to keep models running efficiently on Apple hardware. On macOS, the app adds system-wide text tools for rewriting, summarizing, and correcting text from any application through the Services menu. It also integrates with Siri and Apple Shortcuts, letting users trigger AI tasks by voice or build automations, and supports the x-callback-url specification so it can exchange data with other compatible apps. All processing and conversation history stay on the device. Private LLM is sold as a one-time purchase rather than a subscription, with Family Sharing covering up to six people across a user's Apple devices. It is built by two independent engineers based in the EU without venture funding. Because it depends on on-device hardware for inference, larger models may run more slowly on older devices, and the app is limited to Apple's ecosystem rather than Windows or Android.