Atomic Chat

Atomic Chat is an open-source desktop and mobile app for running large language models entirely on your own machine. The project has 1,824 stars on GitHub and is under active development in 2026. It pairs a friendly chat interface with a local inference engine, so conversations stay private and work fully offline.

Overview

Atomic Chat is built for the local-first AI movement. Download an open-weight model (the app works with GGUF models through llama.cpp and with Apple’s MLX framework on Apple Silicon) and chat with it on your computer or phone, no account and no cloud round-trips. Beyond the chat UI, Atomic Chat exposes an OpenAI-compatible API server on localhost, which makes it a drop-in local backend for other tools, scripts, and agent frameworks that expect the OpenAI API shape. The app is developed by AtomicBot AI and distributed for desktop and mobile platforms. Because inference runs on the device itself, speed and quality depend on the model and hardware you choose, from small models on a phone to large models on a workstation GPU.

Key Features

• Run open-weight LLMs from families such as Qwen, DeepSeek, and Gemma entirely offline
• Optimized inference on Apple Silicon via MLX, plus GGUF/llama.cpp support elsewhere
• OpenAI-compatible local API server (localhost:1337) for use as a drop-in backend
• Desktop and mobile apps with a local-first, private-by-design architecture
• Agent-oriented features, including MCP support for tool-using workflows
• Model management with Hugging Face integration for finding new models

💳 Pricing

Atomic Chat is free and open source. You download the app and the models at no cost; the only requirement is hardware capable of running the model size you choose.

Who Is It For

Atomic Chat is for privacy-conscious users, developers who want a free local OpenAI-compatible endpoint, and anyone who wants to experiment with open-weight models without subscriptions or data leaving their device.