What this does

Installs, switches, deletes, and configures AI models inside the Open WebUI interface without requiring command-line interaction with Ollama. After this guide all model operations are accessible from the WebUI settings panel.

Steps

Open the models settings panel.
- Click the user avatar in the top-right corner
- Select Settings
- Choose Models from the sidebar Expected result: Models page lists all models known to Ollama.
Pull a new model directly from Ollama's registry.
- In the Models panel, locate the Pull a Model input field
- Enter a model reference such as llama3:8b-instruct
- Click Pull Model Expected result: Progress indicator spins until the model downloads.
Set a default model for new chat sessions.
- In Settings > General, find the Default Model dropdown
- Select the preferred model from the list
- Click Save Expected result: Toast notification confirms the setting was saved.
Delete an unneeded model to reclaim disk space.
- In Settings > Models, locate the model row to remove
- Click the trash icon next to the model name
- Confirm the deletion prompt Expected result: The model row disappears from the list.

Verification

curl -s http://localhost:8080/api/v1/models | python3 -m json.tool
# Expected: JSON array containing model names accessible to Open WebUI

Common failures

Model appears in Ollama but not in Open WebUI — The WebUI container may not have connectivity to Ollama. Review OLLAMA_BASE_URL in the compose file.
Pull Model button does not respond — Ollama may not be running. Restart the Ollama container.
Deleted model reappears after refresh — Run docker exec ollama ollama rm <model-name> in addition to the UI deletion.
Model shows "offline" status — Restart Ollama with docker compose restart ollama.
Pull fails with "model not found" — The model reference string is incorrect. Verify with docker exec ollama ollama show <model-ref>.

How to manage models in Open WebUI

What this does

Steps

Verification

Common failures

Related guides