Private AI chatbot that runs open-source LLMs entirely on your iPhone, iPad, or Mac. One-time App Store purchase, no subscription, no account, no cloud. Uses OmniQuant and GPTQ quantization (not RTN like llama.cpp/MLX-based competitors) to maintain text quality while reducing inference cost. Ships 140+ quantized models including uncensored/abliterated variants (Llama, Qwen, DeepSeek, Gemma, Phi). The app itself applies no extra moderation beyond the model’s built-in constraints. Conversations stay on-device; no analytics or tracking SDKs.