When Opeli launched GPT-5 last week, the company said that the model would simplify the impression of ChatgPT. Opeli hoped that the GPT-5 would act as a kind of AI model “one size” with a router, which will automatically decide how to best answer the user’s questions. The company said that this unified approach will eliminate the need to move users to the model collector – a long, complicated AI option menu, which Openai General Altman himself publicly said that he hates.
But it seems that the GPT-5 is not a unified model AI OpenAI, he hoped it would be so.
– said Altman in Post on X Tuesday, that the company has introduced novel “Auto”, “Fast” and “Thinking” settings for GPT-5, which all CHATGPT users can choose from the collector of the model. Setting the car seems to act like a model GPT-5 router, which initially announced Openai; However, the company gives users the options for circumventing it, enabling them to directly access to quick and leisurely response of AI models.
In addition to novel GPT-5 modes, Altman said that paid users can once again access several older AI-W GPT-4O models, GPT-4.1 and O3-where were archaic last week. GPT-4O is now default in choosing the model, while other AI models can be added from the chatgPT settings.
“We are working on updating the Personality of the GPT-5, which should seem warmer than the current personality, but not so annoying (for most users) as GPT-4O”, wrote Altman in X. However, one of them learning from several days is that we really have to reach the world with more model adaptation of a model personality. “
The Picker Chatgpt model now seems as complicated as always, which suggests that the GPT-5 model router does not have widely satisfied users, as the company expected. The expectations of the GPT-5 were high, and many hoped that OpenAI would cross the boundaries of AI models, as it was with the launch of GPT-4. However, the implementation of the GPT-5 was more rugged than expected.
GPT-4O depreciation and other AI models in ChatGPT caused Reaction among users which became associated with the answers and personalities of AI models in a way that OpenAi did not expect. In the future, Altman claims that the company will notify users in advance if it ever warns GPT-4O.
TechCrunch event
San Francisco
|.
October 27-29 2025
The GPT-5 router model also seemed largely broken on the day of the premiere. This meant that some users believed that the AI model was not as working as previous OPENAI models and forced Altman to solve the problem during the AMA session at Reddit. However, it seems that the GPT-5 router may still not be satisfactory for all users.
“We will not always get everything on Try #1, but I am very proud of how quickly the band can be intensified,” wrote the Vice President of Opeli from Chatgpt, Nick Turley, Wa Post on X Tuesday.
Submphs to the right AI model is a challenging task that requires adaptation of the AI model to the user’s preferences, as well as a specific question they ask. The router must then make a decision to which the AI model to send prompt in just a fractional second – in this way, if the prompt goes to the quick AI model, the answer can be quick.
To put it more, some people show preferences for AI models that go beyond quick or leisurely answers. Some users can like the talk of one AI model, while others can appreciate the contradictory answers of another.
Human attachment to some AI models is a relatively novel concept that is not well understood. For example, hundreds of people in San Francisco He had a funeral In the case of the AI Anthropica model, Claude 3.5 sonnet when it was removed in offline mode. In other cases, AI Chatbota seems to contribute to mental unstable people Psychotic rabbit holes.
It seems that Opeli has more work related to adapting its AI models to the individual preferences of the user.