All Local AI
Pick VRAM for fit estimates
Models Tools Hardware Best Of Compare Calculators Stacks Guides Blog
web UI NOASSERTION linux macos windows

Open WebUI

Self-hosted, multi-user chat interface for Ollama and any OpenAI-compatible backend, with RAG and document upload.

GitHub stars

155,956

License

NOASSERTION

Install

docker

Latest release

Oct 16, 2026

Source on GitHub Documentation Official site

Best for

Giving a household or a team a polished chat interface over a local model, with accounts and shared history.

Model formats

  • GGUF

Hardware behaviour

Runs no inference itself, so it is light. All VRAM decisions belong to the backend it points at, typically Ollama or a llama.cpp server.

Not sure what your card can hold? Run the fit check — the entry point in our listings is the RTX 4060 8GB at 8.00 GB.

Setup

Install Open WebUI

Copy-paste starting points. Always read the project's own instructions before running anything with elevated permissions.

Docker (with Ollama on the host)

Open WebUI

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui ghcr.io/open-webui/open-webui:main

Docker (bundled Ollama)

Open WebUI

docker run -d -p 3000:8080 --gpus=all -v open-webui:/app/backend/data --name open-webui ghcr.io/open-webui/open-webui:ollama

pip

Open WebUI

pip install open-webui && open-webui serve

Quick start

Getting the first model running

Start the container, open http://localhost:3000, and create the first account — that account becomes the administrator. Point it at an existing Ollama instance or any OpenAI-compatible URL and every model that backend exposes appears in the model picker.

Beyond chat it handles document upload with retrieval, per-user conversation history, model presets, and role-based access, which is what separates it from a single-user desktop app.

FAQ

Open WebUI questions people actually ask

Short answers to the problems that come up most often.

Does Open WebUI run models itself?

No. It is a front-end. It talks to Ollama or any OpenAI-compatible server, so the model, the VRAM cost, and the speed all come from that backend.

Why can the container not reach Ollama on my machine?

Inside Docker, localhost means the container. Use host.docker.internal with the --add-host flag shown above, or run both on the same Docker network.

Can several people use it at once?

Yes — it supports multiple accounts with separate histories and role-based permissions. Concurrency limits come from the backend: Ollama serialises heavy requests, while vLLM batches them.

How good is the built-in RAG?

Good enough for personal document search: upload files, they are chunked and embedded, and retrieved chunks are added to the prompt. For a production knowledge base you will still want a purpose-built vector pipeline.

Is it safe to expose on the internet?

Only behind a reverse proxy with TLS and real authentication, and with signup disabled after you create the admin account. It was designed for trusted networks first.

Alternatives

Compared with other web UI options

Same job, different tradeoffs.

web UI

text-generation-webui

Gradio web interface that can drive several inference backends, with LoRA training and extensions built in.

 inference engine

Ollama

One-command local model runner with an OpenAI-compatible API and automatic GPU offload.

 inference engine

llama.cpp

The C/C++ engine that made CPU and mixed CPU/GPU inference practical, and the origin of the GGUF format.

 inference server

vLLM

High-throughput serving engine with PagedAttention and continuous batching, built for concurrent requests.

Next step

A runner is only half the decision. Check which models fit your GPU with the VRAM calculator, then follow a setup guide or pick a ready-made stack.