Description
Unsloth is an open-source desktop application for running and training AI models locally on your own hardware. Free and 100% local, it provides a no-code UI to download models from a built-in hub, chat with them, and train or generate media without relying on the cloud.
The app supports text, image, video, and audio models. It can generate images with models such as FLUX and Z-Image and video with Wan and LTX, and it can transform, inpaint, extend, upscale, and edit images locally. Unsloth connects to coding agents like Claude Code and Codex through an 'unsloth start' command and exposes an OpenAI-compatible API, so existing apps, scripts, and SDKs can talk to local models. It also includes private web search and Deep Research inside the app, secure Bash and Python execution for tool-calling, and the ability to serve local or Colab models over HTTPS via a free Cloudflare tunnel.
Unsloth offers Day Zero support for new model releases, a model hub for discovering and managing quantizations, and documentation covering quickstart, integrations, and running specific models locally. It is available for Windows, macOS (Apple Silicon and Intel), and Linux.
Unsloth's Core Features
Free, open-source, 100% local desktop app
No-code UI to run and train models
Built-in model hub with quantization management
Image and video generation and editing locally
Connect coding agents like Claude Code and Codex
OpenAI-compatible API for existing apps and SDKs
Private in-app web search and Deep Research
Serve local or Colab models over HTTPS via Cloudflare tunnel
Getting Started with Unsloth
Download and install: Get Unsloth for Windows, Mac, or Linux.
Load a model: Download a model from the built-in hub and start chatting in minutes.
Train or generate: Use no-code workflows to train models or generate images and video.
Connect an agent: Run 'unsloth start' to connect Claude Code, Codex, or other tools to your local model.
Unsloth's Use Cases
- Local model training
- Local AI coding agents
- Private media generation
- Self-hosted inference






