Private LLM is a native iOS, iPadOS and macOS app by Numen Technologies Limited that runs open-source language models entirely on-device on Apple hardware, without internet or cloud dependency. It uses an mlc-llm based inference engine plus proprietary OmniQuant and GPTQ quantization to run models such as Llama 3.3, DeepSeek R1 Distill, Qwen3 and Gemma 3 efficiently on iPhones, iPads and Macs. It is sold as a one-time purchase with no subscription and has been available on the App Store since June 2023. It integrates with Siri and Shortcuts but does not offer its own OpenAI-compatible API or server functionality.

Preuves (5)