OpenAI has officially rolled out GPT-5, its newest and most advanced ChatGPT upgrade — but the launch hasn’t been as simple as promised.
The company originally pitched GPT-5 as a “one-model-fits-all” AI, using an intelligent model router to choose the best response for any question automatically. The idea? No more “model picker” — a menu of AI versions that CEO Sam Altman has often called clunky.
But in a surprise twist, the model picker is back.
Now, when you open ChatGPT, you’ll see three options — Auto, Fast, and Thinking. Auto works like the original GPT-5 router, deciding for you. But if you want, you can manually choose a faster model or a slower, more in-depth one. While this gives users more control, it also brings back the complexity OpenAI was hoping to eliminate.
Old Favorites Make a Comeback
Paid ChatGPT users will be happy to see some familiar faces return — GPT-4o, GPT-4.1, and o3. GPT-4o is now in the picker by default, while the others can be enabled in settings.
Altman says GPT-5 is also getting personality tweaks — aiming for warmer, friendlier conversations and fewer “odd quirks” that annoyed users in earlier versions. OpenAI is also working on letting people customise AI personalities in more detail in the future.
Mixed Reactions from the Community
Not everyone is thrilled. When OpenAI temporarily removed GPT-4o and other models during testing, many users were upset — especially those who had grown attached to specific AI “personalities.”
Some even complained that GPT-5’s router didn’t perform well on launch day. Altman responded directly in a Reddit AMA, admitting that routing prompts to the perfect model is harder than it looks — the system has to juggle speed, accuracy, and each user’s personal style preferences.
It’s not just about performance, either. Some users prefer certain models for their tone, level of detail, or even their willingness to give “contrarian” answers. In fact, AI attachment is becoming a real thing — earlier this year, hundreds of fans held an actual funeral for Anthropic’s Claude 3.5 Sonnet model when it was retired.
What’s Next for GPT-5
OpenAI says it’s still improving GPT-5 and the model picker. Nick Turley, VP of ChatGPT, praised the team’s fast updates and promised the experience will keep evolving.
For now, the multiple-choice model menu means users can stick with what they know while also trying the latest features. But GPT-5’s launch shows just how tricky it is to build an AI that works perfectly for everyone.
The company’s ultimate goal? An AI that’s fast, smart, and adapts to each user’s personality without making them dig through menus to find it.
