Eris.
A private AI chat app for iPhone and iPad that runs entirely on-device using Apple's MLX framework, supporting models such as Llama, Qwen, Gemma and a distilled DeepSeek reasoning model with no internet connection required.
Eris is an iOS and iPadOS chat app that runs a large language model entirely on-device with Apple's MLX framework, so conversations and data never leave the device and the app works fully offline once a model is downloaded. It supports several quantized general-purpose models, including Llama 3.2, Qwen 2.5, Gemma 2 and Phi 3.5 Mini, as well as reasoning-focused models such as a distilled DeepSeek-R1 variant.
The interface is built natively with SwiftUI and includes syntax highlighting for more than 100 languages in code blocks, full Markdown rendering for formatted text, lists and tables, and both light and dark themes.
It requires an iPhone with an A13 Bionic chip or newer and at least 6 GB of RAM, or an iPad with an M1 chip or newer, running iOS or iPadOS 17.6 or later, with 2 to 8 GB of free storage needed per downloaded model. Building it from source requires an Apple Silicon Mac and a physical iPhone or iPad for testing, since MLX needs real Metal 3 hardware acceleration that the iOS Simulator cannot provide.