unsloth
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
What is unsloth?
Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Overview of unsloth
What it is
Unsloth is a local app for running and training LLMs and diffusion models. The README describes it as the first desktop app to run and train models. It is available in three forms: Unsloth Desktop (the desktop app), Unsloth Studio (the web UI), and Unsloth Core (the code-based version). It works on Windows, Linux, WSL and macOS, and supports multi-GPU setups, NVIDIA, AMD and Intel GPUs, CPUs, and the Vulkan backend.
Key features
- Run and train LLMs, MLX, GGUF, diffusion, decision, embedding and audio models, including Qwen3.8, GLM-5.3-Flash, Kimi K3, Qwen-Image-2.1, MiniMax-H3, DeepSeek-V4 and Gemma 4.
- Agents and tools: use local models with Claude Code, Codex and MCP, including tool calling and code execution.
- Search and RAG: private web search, deep research, auto-compaction (rolling context window) and RAG.
- Fine-tuning: train LLMs, diffusion, decision, TTS and embedding models 2× faster with 70% less VRAM with no accuracy loss, per the README.
- Complete support for reinforcement learning, LoRA, QLoRA, full fine tuning, pretraining, RL, GRPO, DPO and FP8.
- Export and deploy models in GGUF, NVFP4, FP8 and other formats, and serve models through an OpenAI compatible API.
- Remote and LAN access, including a Cloudflare HTTPS link; build datasets from PDFs, CSVs, DOCX files and more with Data Recipes.
Use cases
- Running local models and connecting them to agents such as Claude Code, Codex, Visual Studio Code, OpenCode and others through
unsloth start. - Fine-tuning or pretraining LLMs, diffusion, decision, TTS and embedding models.
- Running and training image and video diffusion or multimodal models.
- Private search, deep research and RAG with local models.
- Serving models over an OpenAI compatible API, or reaching them from other devices on the LAN or remotely.
Getting started
macOS, Linux, WSL:bashcurl -fsSL https://unsloth.ai/install.sh | shWindows:powershellirm https://unsloth.ai/install.ps1 | iexHomebrew:bashbrew install --cask unslothThen launch:bashunsloth studio
License and pricing
Apache-2.0, per the repository metadata. The project is listed as open source; the README does not state pricing.
Based on the project's public README and repository metadata.