LM Mini

关于这款应用

LM Mini is a privacy-focused client application for iOS and macOS that enables users to interact with large language models locally or remotely. Developed by Sultan Haider, the app serves as a unified interface for running open-source AI models directly on hardware owned by the user, eliminating reliance on third-party cloud servers for core functionality.

The application offers three primary modes of operation. First, it supports on-device execution, allowing users to download and run compatible models offline without an internet connection. Second, it provides a connected mode via Wi-Fi or USB to interface with desktop applications such as LM Studio, Ollama, Jan AI, and oMLX. Third, it features LM Mini Connect, which enables secure remote access to local setups from anywhere using user-managed encryption keys.

Key capabilities include downloading and managing models directly from the phone, maintaining local conversation history with folder organization, and tracking context usage. The app supports advanced features such as group chats with up to five models simultaneously, voice interaction via Kokoro, and image generation through ComfyUI and Automatic1111 integrations. Additional tools include web search via SearXNG, semantic search across chats, memory retention for user details, and background generation processes. Users can also import existing chats from LM Studio, utilize optional APIs (including OpenAI, Mistral, and DeepSeek), and employ MCP tool calling. The interface includes dark mode, customizable themes, and encrypted cloud backup options. While performance varies based on device hardware, the app prioritizes user control over data privacy and offline accessibility.