---
title: "Model Lifecycle"
description: "How AI models are introduced and retired in meinGPT, and what you can do as an admin"
canonical_url: "https://meingpt.com/en/docs/admin-guide/model-lifecycle"
language: en
---

# Model Lifecycle

Providers like OpenAI, Anthropic, or Google constantly release new models and retire older ones. meinGPT gives you a predictable process for that: clear phases, timely notice, and a way to switch off a model yourself before we have to.

## The phases of a model

- **Preview** — An early version of a model that can be shut down or replaced at any time. You can try the model in meinGPT, but it shouldn't be used for business-critical processes yet.
- **Active** — The normal state. Fully supported, no restrictions.
- **Retiring soon** — The model is on notice. It still works, but we recommend switching to a successor model. Existing workflows, assistants, etc. using this model can still be used, but it can no longer be selected for newly created workflows.
- **Retired** — The model is no longer available and no longer shown in meinGPT. Existing assistants, workflows, and chats that used it automatically continue with a replacement model.

## How much notice you get

Before a model is discontinued, we notify the admins of an organization by email — usually with several weeks' notice, plus a reminder shortly before the shutdown.
The exact timing depends on how much notice the provider itself gives us: for a very short-notice announcement from a provider, the timeline can be shorter than for a planned, product-driven replacement.

Normally, a shutdown proceeds as follows:
- **4 weeks before shutdown**: First announcement of the shutdown date by email
- **2 weeks before shutdown**: The model switches to "Retiring soon", and we send a reminder email
- **1 week before shutdown**: Final reminder email about the shutdown

## What you can do as an admin

In your workspace settings under **AI Models**, you'll see the "Retiring soon" notice next to each affected model.
You don't have to wait for the shutdown — you can switch at any time:

Open your workspace settings under **AI Models**.

Find the model you want to replace and disable it. If it's still used by assistants or workflows in your organization, the "Disable model" dialog opens — meinGPT shows you how many assistants and workflows are affected.

In that case, choosing a replacement model is required before you can continue.

After confirming, all affected assistants and workflows are immediately switched to the replacement model you chose. This action cannot be undone.

You can't disable a model that's your default model or the only enabled model in your organization — set another model as default, or enable another model first.

This lets you decide for yourself when and what to switch to, instead of waiting for the regular shutdown.

## What happens to your assistants and workflows

When we discontinue a model, the same mechanism runs automatically for you: every assistant, workflow, and workflow step still using the model is switched to the recommended replacement model. Existing chats remain viewable.

## All current models