Hi,
I want to raise a concern about how major changes to ChatGPT are communicated to paying users.
I only discovered yesterday that o3 had been permanently retired from ChatGPT. I understand now that OpenAI announced it beforehand through release notes and official channels, but I was actively using the model and had absolutely no idea it was about to disappear. No email, no meaningful in-app notification. Nothing.
From my perspective, it was literally there one night and gone the next morning.
I find that pretty ridiculous.
The issue isn’t that the model was retired. I completely understand that models evolve and products change. What concerns me is that continuity has quietly become part of the ChatGPT product, but the transition process hasn’t caught up with that reality. My issue is how the transition was handled.
Publishing release notes is not the same thing as notifying the people who are actually affected.
Release notes are passive. Notifications are active.
Why should paying users have to regularly check release notes or Help Centre articles just to discover that something they’ve been relying on for months is about to disappear?
OpenAI already has the telemetry to know which models users actively use. You don’t need to read anybody’s conversations to know that I’ve spent hundreds of hours using o3. That information already exists.
Because of that, notifying the affected users wasn’t impossible, it was a product decision.
The same thing happened to me today with the usage limits and banked resets. I only found out I even had one because I happened to hit my limit. I then had to go looking for what it was, when I got it and whether it expires.
Again… that isn’t communication.
Models also aren’t interchangeable. Anyone who uses ChatGPT extensively knows that. Different models reason differently, solve problems differently and become part of established workflows.
That is exactly what has happened to me.
I’m currently writing three different long-form creative projects with o3. Those conversations have been developing for months. Because I wasn’t notified, I wasn’t given any opportunity to finish those projects, prepare them for migration or even decide how I wanted to handle the transition.
I’m now having to rebuild everything through the API because o3 still exists there. But the API isn’t a replacement for an established ChatGPT workflow.
Inside ChatGPT my conversations already exist. The history exists. The continuity exists.
Moving to the API means rebuilding infrastructure, reconstructing context, managing conversation history myself and effectively creating my own chat application just to continue using the exact same model.
Those are completely different products.
So I’d like to ask whether temporary access to o3 can be restored for existing users who were actively using it, even if only for a short transition period.
More importantly, I’d like OpenAI to seriously consider introducing a notified grace period for future model retirements.
If somebody has been actively using a model for months, they should receive a direct email and a prominent in-app notification, followed by a reasonable period where that model remains available so they can finish existing work and migrate properly.
Not an announcement buried in release notes three months earlier.
Actual notification to the users affected.
Something as simple as:
“o3 will be retired from ChatGPT on August 26. We’ve noticed you’ve been actively using this model. Your access will remain available until [date] to allow you to complete or migrate your existing work.”
That single email would’ve completely changed this experience.
I’d also encourage OpenAI to consider a Legacy Transition Mode for heavily-used models. Thirty days of continued access, export guidance, migration information and behavioural notes would go a long way towards respecting the investment users have already made.
This request isn’t about stopping innovation or keeping older models forever.
It’s about recognising that, for many of us, these aren’t just interchangeable AI models anymore. They’re environments where months or years of creative, professional and research work live. That continuity has value.
If restoring o3 isn’t possible, I understand. But I’d appreciate knowing whether temporary access can be enabled on an account-by-account basis for users who specifically request it, particularly when the model itself continues to exist through the API.
OpenAI has built something people increasingly rely on for long-term thinking, writing, software development, research and creative work. When a model becomes part of that process, continuity becomes part of the product.
I’m not asking for old models to remain forever. I’m asking for transitions that respect the investment users have already made.
I hope this is taken as constructive product feedback, because I genuinely believe solving this would improve the experience for a lot of users, not just me.
Thanks,
Kevin