Free & open source · No account required

Text Generation WebUI (oobabooga)

A Gradio-based interface for running local LLMs on Windows — multiple backends, an OpenAI-compatible API, and full control over the model you load.

Get it from GitHub

Opens the official oobabooga repository · Windows 10/11

Maintainer

oobabooga

Cost

Free

Interface

Gradio (browser)

API

OpenAI-compatible

Two ways to install on Windows

FASTEST

Portable build

Zero setup — download, unzip, and run. Works with GGUF (llama.cpp) models out of the box, no Python needed.

View releases →
MORE BACKENDS

One-click installer

Clones the full repo and sets up a self-contained environment supporting Transformers, ExLlamaV2/V3, and more.

Setup guide →

Setting up the one-click installer

  1. Clone or download the repository

    Use git clone or download the ZIP from the green "Code" button on GitHub.

  2. Run start_windows.bat

    This sets up a self-contained Conda environment inside an installer_files folder — your system Python is untouched.

  3. Select your GPU vendor

    The script asks NVIDIA, AMD, Apple Silicon, or CPU-only, then installs the matching PyTorch build automatically.

  4. Open the local address

    Once setup finishes, browse to http://localhost:7860 and load a model from the Model tab.

System requirements

OSWindows 10 or 11 (64-bit)
GPUNVIDIA recommended; AMD/CPU-only supported with reduced speed
RAM16 GB recommended for 7B-class models
Disk5–15 GB for the app and environment, plus model size
Good to know: This is an independent overview, not affiliated with oobabooga. Both links above point to the real GitHub repository — this page never hosts or serves the installer itself. On Windows, AMD GPU support is more limited than on Linux; WSL is worth considering for AMD cards.