Unsloth AI

Unsloth is an open source project for running and training AI models locally. Its current headline product is Unsloth Desktop, described as the first desktop app to run and train models that is open source, free and fully local. It is offered for Mac, Windows and Linux as a downloadable installer, alongside a Models section, a blog and documentation. The desktop app covers text, image, video and audio models. It can generate and edit images with models such as Qwen Image, FLUX and Z Image plus LoRA adapters, and generate video with Wan and LTX. A built in model hub helps find and download a suitable quantization for the user's hardware, and the vendor says new model releases are supported on day zero. An unsloth start command connects agents such as Claude Code and Codex to local models, and an OpenAI compatible API lets existing apps and SDKs use them too. Other features include private web search with a Deep Research mode that writes cited reports, self healing tool calls, a sandbox where models can run Bash and Python, and a free Cloudflare tunnel for reaching a local or Colab model over HTTPS. The pricing page lists three tiers. Free is the open source version, Pro promises faster training with lower memory use and Enterprise adds multi node support, and both paid tiers are offered through a contact form without published prices. It is used by machine learning engineers, researchers and hobbyists who want to fine tune and run models on their own machines.