LLM Hub is an open-source mobile application for Android and iOS that runs large language models locally on device with CPU/GPU/NPU acceleration. It supports multiple LLM formats including Gemma, Llama, Mistral, Phi4, and Granite, combined with Stable Diffusion for image generation. The app provides RAG-backed memory, multimodal chat, real-time HTML preview for code generation, and offline translation and transcription capabilities.