Atomic-Chat

Atomic-Chat is an agent framework - Local AI app and inference engine for agents. Run open-weight LLMs locally — private, 100% offline on your computer.

Overview

Local AI app and inference engine for agents. Run open-weight LLMs locally — private, on your machine.

or grab any build from atomic.chat · GitHub Releases — latest: v1.1.95

Atomic Chat is built by a small team and a handful of community contributors. Pull requests welcome — see CONTRIBUTING.md for how to get started.

• Run open-weight LLMs locally from HuggingFace — Llama, Gemma, Qwen, Mistral, Phi, and others - Multi-Token Prediction (MTP) speculative decoding — 30–70% throughput boost on supported models, up to 3× on Gemma 4 - DFlash block-diffusion decoding — up to 6× faster on Qwen 3.6, Gemma 4, Kimi K2.5 - Flash Attention toggle ('on' / 'off' / 'auto') - Automatic reasoning-context tracking for chain-of-thought models - Auto context-window expansion with overflow notifications - EAGLE-3 speculative decoding for Gemma 4 on Apple Silicon (MLX) - MTP on MLX for Qwen 3.5 / 3.6 and DeepSeek V4 - TurboQuant KV cache ('turbo3' / 'turbo4') on llama.cpp — now on Windows & Linux too, not just macOS: up to ~4.3× smaller KV cache footprint, CPU and GPU (CUDA / Vulkan) - TurboQuant KV cache on MLX-VLM — smaller memory footprint via RHT-correct fast paths

Atomic-Chat is open-source, written primarily in TypeScript, with 982 GitHub stars under the Other license. The latest release is v1.1.119 (2026-06-18).

Key capabilities

From the project's documentation:

• Run open-weight LLMs locally from HuggingFace — Llama, Gemma, Qwen, Mistral, Phi, and others
• Multi-Token Prediction (MTP) speculative decoding — 30–70% throughput boost on supported models, up to 3× on Gemma 4
• DFlash block-diffusion decoding — up to 6× faster on Qwen 3.6, Gemma 4, Kimi K2.5
• Flash Attention toggle (on / off / auto)
• Automatic reasoning-context tracking for chain-of-thought models
• Auto context-window expansion with overflow notifications

How it fits a local-AI stack

Atomic-Chat runs on your own hardware, so pair it with a model and a GPU sized to your needs. Use the VRAM calculator to pick a model that fits your card, and see what you can run for hardware guidance. Related agent frameworks in the directory:

• Agno
• "ai-agents-from-scratch"
• "aider"
• "airbyte"

Sources

• Source code & docs: AtomicBot-ai/Atomic-Chat
• Official website: https://atomic.chat

Stats from GitHub, 2026-06-27.