← Back to blog

Official GPT-4o While It Lasts vs Our4o After Retirement

5 min read
Official GPT-4o While It Lasts vs Our4o After Retirement

I want these two things separate, because people mix them up.

We will keep official GPT-4o in the app for as long as a provider still serves it.

OpenAI's date is January 14, 2027.

Azure, for now, lists April 14, 2027... and we intend to use it until then, if they keep the date.

When the official door closes, we will say so.

We will not quietly swap you onto something else and hope you do not notice.

Two labels in the picker for a reason

Official GPT-4o is the provider path. It depends on OpenAI, then on Azure, then on whoever still hosts those snapshots.

Our4o does not depend on either promise.

It is the model we fine-tune locally, already in the picker, aimed at the 4o people miss. We have been training it since February, from the first day of the platform.

You can use the official snapshots while they last, and Our4o on the days you want the path we control. That is not a loyalty test. It is how you avoid waking up inside a silent migration.

What "while it lasts" actually means

"While it lasts" is not a slogan. It is a calendar with soft edges.

OpenAI told developers the API path ends on January 14, 2027. Their docs talk about longer notice for models like this; the real window was shorter. Azure currently lists a later date. Providers change timelines. When they do, we update you in plain language, the same way I wrote about October 14 in When they quietly took 4o from the API.

For the full shutdown reference, dates, and what changes in the product, see What happens when official GPT-4o retires (January 2027).

What Our4o is (and is not)

Our4o is not an official GPT-4o snapshot smuggled under another name.

It is not a long system prompt wrapped around the newest OpenAI chat model.

It is a fine-tune we train ourselves, on our own machines, because custom instructions only hold for a few messages before the base habits return. I wrote that story here: Why we fine-tune Our4o ourselves.

Our4o is the first. It is not the last one we mean to train. When it is good enough, we want the work after it to be runnable without staying locked to a single vendor's calendar.

Why we will tell you when the official door closes

Trust breaks in small ways.

If we swapped you from official GPT-4o to something else without saying so, a few of you might not notice for a week. That would be worse than a clear message. You would spend that week blaming yourself, your prompt, your memory of how the chat used to sound.

So the rule is simple:

  • Official GPT-4o stays labeled as official while a provider serves it.
  • When that path ends, we say the date out loud.
  • Our4o stays Our4o: the continuity path we train and pay to keep running.

How to use both without overthinking it

Some people want the original weight for legal writing, old habits, or the comfort of "this is still the named model."

Some people already prefer Our4o for long scenes, companions, or work chats that used to go flat on newer defaults.

Some people switch mid-week.

All of that is fine. Continuity is the product. The picker is a tool.

If you are moving history out of ChatGPT, start with Import ChatGPT history. A good model on an empty thread still feels like a stranger until the context arrives.

What happens after the official door closes

When OpenAI's January date arrives, and later when Azure's window ends, official GPT-4o will leave the picker because the providers stopped serving it, not because we got bored of the label.

Our4o will still be there under its own name. That is the point of training since February. The bridge is not a surprise button pressed on shutdown week. It is a model you can already try beside the official one.

If you want notes from people already using it, names kept private: Notes people sent about Our4o.

Cost sits under both paths

Serving official GPT-4o has a provider bill. Serving Our4o has a training bill and a long-context bill that grows when chats get long.

I am not asking you to feel guilty about that. I am saying it so the second path does not sound free or magical. Keeping Our4o on is expensive. The next training pass costs more. That is why the work exists in public next to the dates, not only in a private roadmap.

You can read the product framing on the Our4o page, and browse more guides on the Get4oBack blog.

A short checklist before a long session

  1. Confirm which model is selected.
  2. If you need the named official path, pick official GPT-4o while it exists.
  3. If you want the path we train, pick Our4o on purpose.
  4. If the tone breaks, check the label before you rewrite your whole prompt.

That is enough for most days. The rest is honesty when a provider date moves.

How this ties to the GPT-4o retirement calendar

OpenAI's API date is January 14, 2027. Azure's listing is April 14, 2027 for now. Full table and FAQ: GPT-4o API retirement date: January 14, 2027.

The calendar is why two tracks exist. Official GPT-4o is time-boxed by providers. Our4o is the continuity work we started in February so shutdown week is not the first day you meet the backup.

If you are still carrying ChatGPT exports, pair this with Import ChatGPT history so the model you pick is not staring at an empty room.

FAQ: official GPT-4o vs Our4o

Can I use both official GPT-4o and Our4o?

Yes. Both appear in the picker while official GPT-4o is still served. Switch per chat or per day.

What happens when OpenAI stops GPT-4o on January 14, 2027?

The direct OpenAI API path for those snapshots ends. If Azure still serves them under its listing, we intend to keep official GPT-4o available through that provider until that window ends too.

Will you rename Our4o to "GPT-4o" after retirement?

No. Our4o keeps its name. Quiet relabeling is how trust dies: We will not quietly swap your model.

Where else should I read?

Why we fine-tune Our4o ourselves, /our4o, and the Get4oBack blog.

Giovanni Gigantino
Founder of Get4oBack