Atomic-Chat
About Atomic-Chat
Atomic Chat is a local AI application and inference engine designed to run open-weight large language models (LLMs) on a user's computer. It provides a private, offline environment for interacting with AI models without relying on cloud services.
The software supports multiple inference engines, including a custom fork of llama.cpp with TurboQuant optimizations, the upstream llama.cpp, and MLX-VLM for Apple Silicon. It offers an OpenAI-compatible API server, allowing integration with various tools and agents that support the OpenAI API format.
Key features
- Run open-weight LLMs locally from HuggingFace
- Multi-Token Prediction (MTP) speculative decoding for performance boosts
- DFlash block-diffusion decoding for faster inference on specific models
- Flash Attention toggle for optimized processing
- OpenAI-compatible local API server
- Support for multiple inference engines including TurboQuant-optimized llama.cpp and MLX-VLM
- One-click agent launch for coding assistants
- Projects with conversation tree view
Download v1.1.154
Jul 14, 2026WindowsWindows
macOSmacOS
Frequently asked questions
- Is Atomic-Chat free?
- Yes. Atomic-Chat is free and open-source software, distributed under the NOASSERTION license.
- Which platforms does Atomic-Chat support?
- Atomic-Chat provides downloads for Windows, macOS, Linux.
- What is the latest version of Atomic-Chat?
- The latest version is v1.1.154, released on Jul 14, 2026.
- What platforms does Atomic Chat support?
- Atomic Chat is available for macOS (Universal), Windows (x64), and Linux (AppImage). Mobile versions are also available for iOS and Android.
- How does the local API work?
- Atomic Chat runs an OpenAI-compatible server at http://localhost:1337/v1, which can be used as a drop-in replacement for the OpenAI SDK. This allows any tool that supports the OpenAI API to connect to locally running models.
- What inference engines does Atomic Chat use?
- The software uses three main engines: a custom TurboQuant-optimized llama.cpp fork, the upstream llama.cpp, and MLX-VLM for Apple Silicon devices. These are exposed through a unified API interface.
- Is internet connection required to use Atomic Chat?
- No, Atomic Chat is designed to work completely offline when using local models. However, it also supports cloud-based models if internet connection is available and desired.
More AI & LLM Tools
ollama
ollama
Ollama provides access to a variety of AI and large language models, including Kimi-K2.6, GLM-5.1, MiniMax, DeepSeek, gpt-oss, Qwen, and Gemma. It simplifies the process of setting up and using these models for various applications.
dify
langgenius
Dify is a production-ready platform for developing agentic workflows. It supports AI and LLM tools for streamlined workflow creation.
anything-llm
Mintplex-Labs
AnythingLLM is a local-first AI tool for running language models independently. It provides a self-hosted alternative to cloud-based solutions for users seeking full control over their AI agents.