TextGen (text-generation-webui) Download: Local LLM App

For years, “oobabooga” has been the go-to web interface for people who like to tinker with local language models. The project now lives at github.com/oobabooga/textgen under the name TextGen and calls itself a desktop app for local LLMs.

What does TextGen do?

• Several backends: llama.cpp, ik_llama.cpp, Transformers, ExLlamaV3 and TensorRT-LLM, switchable without a restart.
• Chat modes: Instruct, chat-instruct and chat, with Jinja2 prompt templates.
• Files and images: Attach text, PDF and .docx files, and use vision models.
• API: An OpenAI- and Anthropic-compatible API with tool calling, plus MCP server support.
• Extras: LoRA training, an image generation tab for diffusers models, and extensions for text-to-speech, voice input and translation.
• Privacy: Runs offline, with no telemetry or remote update requests.

Who should use TextGen?

It is for hobbyists and power users who want fine control over models, sampling settings and backends, while still having a friendly interface.

Portable build or full install?

Most people should start with a portable build. Grab the build for your system and GPU from the GitHub releases, unzip it and double-click textgen; a window opens and everything it needs is included.

For extra backends, training, image generation or extensions, use the full installation. Clone the repository and run the start script for your system, then choose your GPU vendor when asked:

git clone https://github.com/oobabooga/textgen
cd textgen
./start_linux.sh

On Windows run start_windows.bat, and on macOS start_macos.sh. Afterwards the interface is at http://127.0.0.1:7860. Put GGUF model files in the user_data/models folder and the app finds them automatically.

Hardware requirements

• Portable builds: Available with CUDA, Vulkan, ROCm or CPU-only support, and run GGUF models.
• Full installation: Needs about 10 GB of disk space, downloads PyTorch, and adds support for 16-bit Transformers and EXL3 models.
• Memory: Depends on the model and context length; the project links a GGUF memory calculator to estimate it.

TextGen is free and open source under the AGPL-3.0 license.