Verified project record

unslothai/unsloth

A library and local web UI for running and fine-tuning language, audio, embedding, and vision models on your own hardware, designed to reduce speed and VRAM requirements. It provides automated installation, document-based dataset creation, and multiple export formats, though server-side tools run with the user's permissions by default.

Project overview

The project reduces the compute barrier to local model training and inference, enabling users to run and customize models on consumer hardware while offering a no-code visual interface for dataset preparation.

Project type: Model Development · Model Runtime · Evaluation & Observability
Use cases: Audio & Speech
Deployment: Refer to project documentation
License: Apache-2.0

Best for

AI engineers and data teams running, training, or fine-tuning models locally who need to operate within constrained GPU memory limits.

Key capabilities

Run inference on various model formats including GGUF, LoRA adapters, and safetensors.
Auto-create datasets from sources such as PDF, CSV, and DOCX, and edit data using a visual-node workflow.
Save or export trained or merged models to formats including GGUF, 16-bit safetensors, and other specified formats.
Monitor training live, track loss and GPU usage, and customize graphs.

Limitations and risks

Server-side tools run as your user and are enabled by default; anyone who can reach the server with the API key can execute code.

Getting started

Installation is rated as easy due to an automated install script for Windows, macOS, Linux, and WSL. The first success path involves running the installation script, launching the web UI with `unsloth studio -p 8888`, and accessing it via a browser.

Evidence and sources

GitHub project description: Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.
README: Unsloth can be used in two ways: through **[Unsloth Studio](https://unsloth.ai/docs/new/studio/)**, the web UI, or through **Unsloth Core**, the code-based version.
README: Unsloth Studio lets you run and train models locally.
README: Train and RL **500+ models** up to **2x faster** with up to **70% less VRAM**, with no accuracy loss.
README: Search + download + run models including GGUF, LoRA adapters, safetensors

Reviewed at: 2026-07-17T05:32:21Z

Open original GitHub project