When Openai started the GPT-5 last week, the company said the model would simplify the experience of Chatgpt. Openai hopes that the GPT-5 would serve as a kind of “of a size that fits everyone” AI model with a router who will automatically decide how to better answer user questions. The company said this unified approach would eliminate the need for users to browse its model – a long, complicated menu of AI options that Openai Sam Altman has said publicly hating.
But it seems that GPT-5 is not the unified AI Openai model that hopes it would be.
Said Altman to a Post in x On Tuesday that the company introduced new settings “Auto”, “Fast” and “Thinking” for the GPT-5 that all Chatgpt users can choose from the Picker model. The automatic setting seems to work like the GPT-5 model router initially announced by Openai. However, the company also gives users’ choices to wander it, allowing them to access quickly and slow AI models directly.
Along with the new ways of GPT-5, Altman said that paid users can access many AI Legacy models-including GPT-4O, GPT-4.1 and O3-who were frustrated just last week. The GPT-4O is now on the default pick-up Picker model, while other AI models can be added from Chatgpt settings.
“We are working on an GPT-5 personality that should feel warmer than current personality, but not as annoying (to most users) as GPT-4O,” Altman writes in place of X. “However, learning for us in recent days is really just to reach a world.
Chatgpt’s Picker model seems to be as complex as ever, indicating that the GPT-5 model router does not have universally satisfied users such as the company was hoping. Expectations for the GPT-5 were high, with many hoping that Openai would push the AI models as it had with the onset of GPT-4. However, the development of the GPT-5 was tougher than expected.
The depreciation of GPT-4O and other AI models in Chatgpt caused a reaction between users who had been developed with the answers and personalities of AI models in ways that Openai had not foreseen. In the future, Altman says the company will give users a lot of warning if it ever removes the GPT-4O.
TechCrunch event
Francisco
|
27-29 October 2025
The GPT-5 model router also seems to break a lot on the launch day. This caused some users to feel that the AI model was not as extensive as previous Openai models and forced Altman to deal with the problem at an AMA session on Reddit. However, it seems that the GPT-5 router may not be satisfactory for all users.
“We’re not always going to get everything in Try #1. But I’m very proud of how fast the team can repeat,” wrote Chatgpt’s Openai Vice President Nick Turley in a Post in x Tuesday.
Examing to the right AI model is a difficult task that requires the alignment of an AI model in a user’s preferences, as well as the specific question they ask. Then the router must make a decision for which the AI model to send the prompt in just one second – in this way, if a prompt goes to an AI model that responds quickly, the answer can still be fast.
In general, some people have preferences for AI models that go beyond fast or slow answers. Some users may like the AI model’s deficit state, while others may appreciate another’s contradictory answers.
Human attachment to some AI models is a relatively new idea that is not well understood. For example, hundreds of people in San Francisco recently Held a funeral For Anthropic’s AI model, Claude 3.5 Sonnet, when it was pulled off. In other cases, AI Chatbots seem to contribute to mentally unstable people who go down Psychotic rabbit holes.
It seems that Openai has more work to align AI models to individual users’ preferences.
