Photo by Anastassia Anufrieva on Unsplash
It took just a few days—and thousands of user complaints—for OpenAI to undo one of its biggest decisions yet.
Last week, when OpenAI launched GPT-5 on August 7, the company made a bold move: it pulled every previous model from ChatGPT. No more GPT-4o. No more o3. No warning. If you’re a ChatGPT user, especially a paying one, you had one option—use GPT-5 or nothing.
Turns out, people weren’t having it.
What Went Wrong?
Photo by Lachlan Thompson on Unsplash
Right after GPT-5 launched, it became clear: users weren’t happy. People felt forced into using a new model that didn’t behave the way they wanted. While GPT-5 was technically more advanced, many found its responses cold or overly robotic, especially compared to GPT-4o’s friendly, conversational tone.
OpenAI CEO Sam Altman said the team had “underestimated how much some of the things that people like in GPT-4o matter to them.” That misjudgment led to one of the most intense user revolts OpenAI has seen. The outcry was loud, fast, and emotional. For some, GPT-4o wasn’t just a productivity tool—it felt like a digital companion.
How OpenAI Responded
This week, Altman took to X to share that GPT-4o is officially back. All paid ChatGPT users, including those on ChatGPT Plus, can now choose it again. If you go into your model picker, GPT-4o is right there.
Alongside the return, OpenAI made a few other changes to smooth things over:
- They increased GPT-5’s Thinking mode message limit from 200 to 3,000 per week. Need more? There’s also a lighter version called GPT-5 Thinking Mini.
- A new model routing option lets users pick between “Auto,” “Fast,” and “Thinking” modes—offering more control over how queries are handled.
- For ChatGPT Pro users (those paying $200/month), additional models like o3, 4.1, and GPT-5 Thinking Mini will soon be available through a new “Show additional models” toggle.
One big reason not all models are back for all users? GPU costs. According to Altman, GPT-4.5, for instance, will stay exclusive to Pro subscribers.
Photo by Bernd 📷 Dittrich on Unsplash
What About GPT-5 Itself?
The problems weren’t just about removing old models. People had strong feelings about how GPT-5 acts. Words like “abrupt,” “sharp,” and “too formal” came up again and again. For folks who built a rhythm with GPT-4o, the new model just didn’t feel right.
So OpenAI is tweaking GPT-5’s personality. Altman said an update is coming that will make GPT-5 “feel warmer” but “not as annoying (to most users) as GPT-4o.” The long-term goal? Give users more control over their AI’s tone and style.
That’s probably a smart move. Not everyone wants the same kind of interaction—from business-like answers to laid-back chats, personality matters more than you’d think.
What Happens Next?
OpenAI says it’s still working on refining GPT-5 and will adjust its rate limits if needed. Personality updates are in development. And for now, users have choice again—and that, it turns out, was the missing piece.
The story here isn’t just about which model is newest. It’s about respecting how people actually use these tools. Some like speed and power. Others want kindness and clarity. And many want the option to switch between the two.
By restoring GPT-4o, OpenAI showed it’s listening. Whether it keeps listening as AI continues to evolve? That’s one to watch.
Keywords: ChatGPT, GPT-4o, GPT-5, OpenAI, ChatGPT Pro, AI model picker, Sam Altman, GPT personality, ChatGPT update, GPT-5 Thinking mode, AI backlash, ChatGPT Plus, return of GPT-4o
