AltFreeStack
ESEN
Migration guide

How to migrate from ChatGPT Plus to Open WebUI

Unlike other migrations, there's no {from} export you can re-import into {to} as-is — what actually changes is where the model runs and who sees your conversations, not a data file.

  1. 1

    Export your ChatGPT Plus history as a reference file only

    From Settings → Data Controls → Export Data in ChatGPT Plus you can download a .json of your history — keep it for reference, since there's no way to re-import it directly into Open WebUI.

  2. 2

    Deploy Open WebUI alongside a backend that runs the model

    Open WebUI is just the chat interface: it doesn't run models on its own. The simplest path is Ollama in a separate container (see the docker-compose on its page) — it works without a GPU, though a GPU speeds up responses significantly.

  3. 3

    Download at least one local model with Ollama

    Ollama needs you to pull a model (for example Llama or Mistral) before you can chat from Open WebUI — the model size you pick determines how much RAM/VRAM you need, not Open WebUI itself.

  4. 4

    Or connect Open WebUI to a remote OpenAI-compatible API

    If you don't want to depend on local hardware, Open WebUI also connects to any OpenAI-compatible API (including OpenAI itself) — you lose the privacy of a fully local model, but keep the same interface with commercial-model quality.

  5. 5

    Recreate your saved prompts and custom instructions

    ChatGPT Plus's custom instructions and saved prompts don't migrate automatically — check the .json from step 1 and save them again in Open WebUI.

Before cancelling ChatGPT Plus

Before cancelling ChatGPT Plus, test Open WebUI with 3-5 real tasks you used to solve with it — response quality depends entirely on the model you choose to run (local or remote via an API), it isn't a fixed property of Open WebUI.

Ready to deploy Open WebUI?

Full profile with license, tech stack and ready-to-copy docker-compose.

View Open WebUI's profile

These steps are a general guideline — the exact export menu can vary depending on ChatGPT Plus's plan and version.

Frequently asked questions

Does Open WebUI need a GPU?

It's not required. Open WebUI is just the interface — inference happens in Ollama or another backend. It works on CPU without a GPU, but a GPU speeds up response time significantly.

Can I use Open WebUI without fully local models?

Yes. It also connects to any OpenAI-compatible API (including OpenAI itself), if you prefer commercial-model quality without changing interfaces.

What do I lose compared to ChatGPT Plus?

Response quality depends entirely on the model you choose to run — a small local model won't always match the latest commercial models, though connecting Open WebUI to a commercial API closes that gap a lot without losing the interface or control over your prompts.