Updated: 28 September 2026 · Applies to: Open WebUI 0.11 with Docker and Ollama 0.34
When Open WebUI shows no models or a connection error, it cannot reach the Ollama API. Almost always the cause is that Ollama only listens on 127.0.0.1, while Open WebUI runs in a Docker container that sees the server as a different machine. Follow these checks in order.
1. Check that Ollama runs
systemctl status ollama curl http://127.0.0.1:11434
The answer must be Ollama is running. If not, start it with sudo systemctl start ollama and read journalctl -e -u ollama.
2. Look at the connection setting in Open WebUI
- Log in as administrator, click your avatar and open Settings › Admin › Connections.
- Under Manage Ollama API Connections check the URL. When Open WebUI runs in Docker with the default (bridge) network and Ollama runs on the server itself, use
http://host.docker.internal:11434.
3. Fix it: pick one of two ways
Way A: host networking (nothing is opened to the network). Recreate the container so that it shares the server's network. Your data stays in the volume:
docker rm -f open-webui docker run -d --network=host \ -v open-webui:/app/backend/data \ -e OLLAMA_BASE_URL=http://127.0.0.1:11434 \ -e WEBUI_SECRET_KEY=YOUR-EXISTING-KEY \ --name open-webui --restart always \ ghcr.io/open-webui/open-webui:main
The interface is now on port 8080 instead of 3000. Use the same WEBUI_SECRET_KEY as before, otherwise users are logged out.
Way B: let Ollama listen on more than localhost. Open WebUI's documentation names this as the fix when Ollama only listens on 127.0.0.1: set OLLAMA_HOST=0.0.0.0, make sure the setting is part of the Ollama service, and restart Ollama (How to make Ollama reachable from other machines (OLLAMA_HOST) safely? shows the exact steps). Because this opens port 11434 to the network, also restrict it with a firewall as described there. In the Open WebUI container use --add-host=host.docker.internal:host-gateway so the name host.docker.internal works on Linux.
4. Settings page is slow or the model list will not load
An unreachable connection that you saved earlier can make the page wait for its timeout. To make Open WebUI give up sooner, add -e AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST=3 to the docker run command (the default is 10 seconds), and remove or correct the wrong connection in Settings › Admin › Connections.
5. Firewall
If Ollama is on another server, check that the firewall on that server allows the port from this server: from the Open WebUI server run curl http://OTHER-SERVER:11434. It must answer Ollama is running.
6. Models are missing after a fix
Open WebUI lists the models of the Ollama it is connected to. If you connected to a different Ollama (for example a container with its own empty model folder) the list is empty: run ollama list on the Ollama you want and check the connection URL again.
Frequently asked questions
Which way is better?
Way A, when Ollama and Open WebUI are on the same server, because Ollama stays private. Use Way B only when a container that cannot use host networking needs Ollama, and then restrict access.
Do I lose my chats when I recreate the container?
No. Chats, users and settings are in the open-webui volume, which the new container uses again.
Official documentation: Open WebUI documentation: connection errors.
Want your own private AI without the setup work?
- Private LLM installation: we install and configure Ollama, the GPU driver and a chat interface on your own server.
- GPU dedicated servers: NVIDIA GPU servers for AI inference and training.
- AI implementation services: private LLMs, RAG, n8n automation and API integrations built on your own servers.
Prefer a hand with the setup? Our engineers can do it for you: Hire an Expert, or use our on-demand server management.
