TextGen
TextGen, formerly text-generation-webui, is an AGPL-3.0 desktop app for running language models on your own computer with no telemetry. It switches between five engines, including llama.cpp and ExLlamaV3, without restarting, serves an OpenAI and Anthropic-compatible API to other apps, calls tools such as web search and MCP servers, chats about uploaded PDFs and documents, generates images, and trains LoRA adapters. It runs fully offline.
What does TextGen do?
TextGen, the project long known as text-generation-webui, is "a desktop app for local LLMs. Open source, no telemetry." It runs models through llama.cpp, ik_llama.cpp, Transformers, ExLlamaV3 or TensorRT-LLM, and switches between engines and models without restarting. Its privacy line: "100% offline and private, with zero telemetry, external resources, or remote update requests."
It serves an OpenAI and Anthropic-compatible API with chat, completions and messages endpoints and tool calling, so other apps can use it as a drop-in local replacement. Tools include web search, page fetching and maths, each a single Python file, and MCP servers are supported. You can upload text, PDF and DOCX files to discuss, attach images for vision models, generate images with diffusers models, and fine-tune LoRA adapters on chat or raw text. Extensions add speech and translation. A multi-user mode with password sign-in exists for small trusted teams.
Key facts
- Licence: AGPL-3.0, unmodified.
- Engines: llama.cpp, ik_llama.cpp, Transformers, ExLlamaV3 and TensorRT-LLM, switchable without restarting.
- An OpenAI and Anthropic-compatible API with chat, completions and messages endpoints and tool calling.
- Tools: web search, page fetching and maths, each a single .py file; MCP servers supported.
- File chat for text, PDF and DOCX; image input for vision models; image generation with diffusers models; LoRA training.
- No telemetry, external resources or remote update requests, per the README.
- Install: portable builds (download, unzip, run), pip in a Python 3.9+ venv, a one-click installer, or Docker Compose for NVIDIA, AMD, Intel or CPU; the full install needs about 10 GB of disk.
- The repository moved from oobabooga/text-generation-webui to oobabooga/textgen. Latest tagged release when read: v4.9 on 2026-05-20; 47,683 stars on 2026-09-17.
What does it replace, and where does it fall short?
TextGen replaces ChatGPT for someone who wants the most control over local models: which engine, which quantisation, which extensions, and the option to train. Its API also replaces the OpenAI API for other programs on the machine, which is why it appears on the open-source alternatives to the OpenAI API page.
Where it falls short: it is a tool for people comfortable with model files and engines; the portable build covers GGUF models only, and other formats need the full install at about 10 GB. Multi-user mode does not save chat histories, and the README says it is best for small trusted teams. For a simpler desktop app, Jan.
How does TextGen run?
Download a portable build, unzip and run it; it includes everything for GGUF models through llama.cpp. For all engines, use the one-click installer for Windows, Linux or macOS, which sets up Conda, or a manual install with Python and PyTorch, or Docker Compose with files for NVIDIA, AMD, Intel or CPU. The interface opens at 127.0.0.1:7860. No hosted version is described.
Who is TextGen for?
A technical user who wants every knob for running, comparing and training local models, and an OpenAI-compatible API for other tools. Someone who wants a chat app that just works should start with Jan or Chatbox.
What limits does the README state?
From the README: Transformers and ExLlamaV3 formats "require the full installation (not the portable build)"; the full install "requires ~10GB disk space"; in multi-user mode "chat histories are not saved or automatically loaded. Best suited for small trusted teams."
Questions people ask
Is TextGen the same as text-generation-webui?
Yes. The repository moved from oobabooga/text-generation-webui to oobabooga/textgen and the README now calls the product TextGen. The old address redirects to the new one. It is the same AGPL-3.0 project, with the same maintainer and the same install routes.
Does TextGen send any data out?
No, by its README: "100% offline and private, with zero telemetry, external resources, or remote update requests." Models run locally through its five engines. Web search is a tool you choose to enable.
Can other apps use TextGen as an OpenAI API?
Yes. It serves an OpenAI and Anthropic-compatible API with chat, completions and messages endpoints and tool calling, which the README describes as a local drop-in replacement for those APIs. See the open-source alternatives to the OpenAI API page for the dedicated servers.
How does TextGen compare with ChatGPT?
TextGen runs open models on your own hardware with full control over engines, tools and training, offline. ChatGPT offers OpenAI's models with no setup. The open-source alternatives to ChatGPT page compares TextGen with the simpler desktop apps and the team platforms.
What hardware does TextGen support?
NVIDIA through CUDA, AMD through ROCm, Vulkan, Apple Silicon and Intel Macs, Intel GPUs through Docker, and CPU only. Portable builds cover GGUF models; the full install supports every engine and needs about 10 GB of disk space.
Sources
- TextGen README and AGPL-3.0 LICENSE: github.com/oobabooga/textgen, read 2026-09-17.
Compared with the others
On the open-source alternatives to ChatGPT page, TextGen is the pick for the most control over local models. AGPL-3.0, five engines including llama.cpp and ExLlamaV3, an OpenAI and Anthropic-compatible API, tools, file chat, image generation and LoRA training, fully offline.
Also on that page: Open WebUI for a private chatgpt for a whole team, LibreChat for every provider in one mit platform, AnythingLLM for chatting with your documents.
On the open-source alternatives to the OpenAI API page, TextGen is the pick for a local api with every engine. AGPL-3.0, five engines and an OpenAI and Anthropic-compatible API with tool calling, fully offline.
Also on that page: Ollama for the quickest local start, LocalAI for every modality behind openai-compatible apis, llama.cpp for the engine on any hardware.
More chat interfaces
A self-hosted chat platform for Ollama and OpenAI-compatible models with document search over 9 vector databases, web search, image generation, roles and single sign-on, under a BSD-3 licence with an added branding clause.
LobeHub, formerly LobeChat, is an agent workspace with an agent builder, agent groups, scheduled runs, editable memory and a library of over 10,000 tools and MCP plugins, under a community licence that needs a paid licence to distribute a modified version.
An MIT all-in-one AI app from Mintplex Labs for chatting with your documents with citations, running agents with MCP, and using any of 39 model providers including local llama.cpp and Ollama, as a desktop app or a multi-user Docker server.
An AGPL-3.0 desktop client for Windows, macOS and Linux that connects OpenAI, Gemini, Anthropic and local Ollama or LM Studio models, with 300+ preset assistants, side-by-side model chats, file support and MCP servers.
An Apache-2.0 desktop app from Menlo Research that downloads and runs open models from Hugging Face on your own computer, connects to OpenAI, Anthropic and others when you want, serves a local OpenAI-compatible API, and supports MCP.
Added September 17, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 17, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/oobabooga/textgen.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.