OnDevice LLM – Offline AI Chat

关于这款应用

OnDevice LLM is an iOS application that executes large language models locally on the device, ensuring that all data processing occurs offline without connecting to external servers. The app supports a variety of open-source models, including GGUF formats like Llama 3.2 and Qwen, as well as Apple Silicon-native MLX models such as Ternary Bonsai. Key features include a persistent personal memory that stores facts locally, voice input via on-device speech recognition, and the ability to attach PDFs or images for local analysis. Users can customize system prompts and load models from HuggingFace URLs with a Pro upgrade. The interface includes quick-start personas for tasks like coding or translation, and mini-apps like calculators or games that run within the chat environment. Designed for privacy-conscious users, it requires no account creation, email, or tracking SDKs, making it suitable for professionals handling sensitive information or travelers in areas with limited connectivity.