Unsloth AI

Unsloth AI is a coding tool. Unsloth Desktop is a free, open-source, no-code app for running and training text, image, video and audio models locally on Mac, Windows and Linux, with agent connections, an OpenAI-compatible API and built-in web search.

Unsloth Desktop is an open-source application for running and training open AI models on local hardware. It packages text, image, video and audio models into a single no-code interface, so work stays on the user's own machine instead of passing through a hosted service. Installers are offered for Mac (Apple Silicon and Intel), Windows 10 and later, and Ubuntu-compatible Linux distributions. The workflow starts with a built-in model hub, where users browse, manage and download a quantization suited to their device, then chat with the model within minutes. The same app handles no-code training workflows and media generation. Image tools cover creating, transforming, inpainting, extending, upscaling and editing pictures with models such as Qwen-Image-2.1, FLUX and Z-Image plus LoRA adapters, while video generation uses Wan and LTX. The project says it offers day-zero support for new releases from families like Qwen, Gemma, GLM and NVIDIA. Several features target agent and developer use. The `unsloth start` command connects Claude Code, Codex and similar agents to a local model, so the model can be swapped without changing the agent workflow. An OpenAI-compatible API lets existing apps, scripts and SDKs talk to local models. Models can run Bash and Python in a sandbox, and the app advertises self-healing tool calls that detect, repair and retry failures. Web search and a Deep Research mode that produces cited reports are built in, and local or Colab models can be served over HTTPS through a free Cloudflare tunnel for remote monitoring. Unsloth Desktop suits developers, researchers and hobbyists who own capable GPUs or Apple hardware and want privacy and no per-token fees. It sits in the same space as other local runners and chat front ends, but combines fine-tuning, media generation and agent connectivity in one package. Performance depends on the user's hardware, and larger models and video workloads will call for strong GPUs.