Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- OpenAI is bidding adieu to 4o, a favorite model among many users.
- Chats with 4o will still be accessible after that date.
- The company says only 1% of its users still engage with 4o.
It's a sad day for many ChatGPT users.
GPT-4o, the model that's become a favorite for many people because of its warm and friendly conversational tone, will be retired on Feb. 13, OpenAI announced Thursday. That means it'll no longer be accessible to paid users via the model selector menu in ChatGPT, though it'll still be available through OpenAI's API for the foreseeable future. ChatGPT's free users — which comprise around 90% of OpenAI's userbase — have not had access to 4o since OpenAI debuted GPT-5 in August.
(Disclosure: Ziff Davis, ZDNET's parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
Also: Here are all the GPT-5 updates OpenAI has rolled out since launch
GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and Thinking will also be discontinued starting Feb. 13. But OpenAI is treading extra carefully in the case of 4o, after it faced user blowback when the model was briefly removed after the release of GPT-5.
“We know that losing access to GPT‑4o will feel frustrating for some users, and we didn't make this decision lightly,” the company wrote in its announcement. “Retiring models is never easy, but it allows us to focus on improving the models most people use today.”
What to expect
The good news is that for most ChatGPT users, the transition should feel pretty seamless — at least in the most practical terms.
If you're a Plus or Pro subscriber, regardless of which models you've been using, all of your preexisting conversations with ChatGPT will still be accessible after the Feb. 13 retirement date. You'll be able to continue them, just using different models. Free users will still use GPT-5 as their default model, so for all intents and purposes, the change won't be noticeable for them.
Existing GPTs based on 4o or any of the other soon-to-be-retired models will be switched to GPT-5.2 on Feb. 13, according to OpenAI's website. ChatGPT's voice mode and image-generating capabilities won't be affected by the switch, since they both currently use a model that's different from (although similar to) 4o.
Vibe shift
The most noticeable change will be for those paying users who still use 4o because they prefer its tone and conversational style of 4o over those of newer models.
This became a major flashpoint for the company when 4o was initially discontinued following the August GPT-5 release. The company was quick to reinstate it for some paid users, and it promised to heed the negative feedback moving forward. Both GPT-5.1 and 5.2 came with various “personalizable presets,” including “professional,” “candid,” and “quirky,” the idea being to make the models more adjustable to individual user preferences.
Also: I put GPT-5.2 through a 14-round test, and the AI model raised some serious questions
In its Thursday announcement, OpenAI wrote that only around 1% of ChatGPT users still engage with 4o. Still, some of those 900,000 or so paying customers clearly resent OpenAI's decision to retire the model.
A couple of them have argued that the 1% figure is misleading, since it factored in the much more massive swathe of free users who don't have access to 4o. “I think that percentage would be vastly different if you look at [subscribed users] only,” one person wrote in the OpenAI subreddit on Thursday night.
Another was already trying to rally support to reverse the decision. “Time to go to change.org and start filling out petitions again,” they wrote. “We brought 4o back last time. We'll bring it back again.”
