Text Generation Web UI (often called « oobabooga ») is one of the leading web interfaces for running large language models locally. It brings together in a single tool model loading, a chat interface, a notebook mode, and an API, all self-hostable and offline capable.

Its key strength is the versatility of inference engines: it supports multiple backends (llama.cpp, Transformers, ExLlamaV3) and many quantized model formats (GGUF, EXL2), allowing you to load virtually any open-weight model and switch between them on the fly.

It exposes an OpenAI-compatible API to connect your applications, offers an extension system, character and prompt template management, and a simple Docker deployment — ideal for experimenting and serving your own models without relying on a cloud service.