LocalEngine: Think Tank

关于这款应用

LocalEngine is a privacy-focused application that runs open-source AI language models directly on iPhone and iPad hardware. By utilizing a Metal-accelerated llama.cpp engine built in pure Swift, it delivers low-latency, offline inference without sending data to external servers. The app supports multi-turn text conversations with real-time token streaming and includes vision capabilities for analyzing attached images, provided the selected model supports multimodal input.

Users can browse a curated catalog of open models, including Qwen3.5 and Gemma 4 variants. Smaller models (0.8B/2B for Qwen, E2B for Gemma) are available for free, while larger, more capable models require a one-time Pro upgrade. The interface allows users to adjust generation parameters like temperature and max tokens. LocalEngine requires no account creation, sign-in, telemetry, or analytics, ensuring complete user privacy. It supports iOS/iPadOS 17.0+ and features a fully localized interface in ten languages.