Only one day after having removed the option to select the old LLM for the benefit of GPT – 5, OPENAI reintroduced GPT – 4O as an option in Chatgpt. The decision to define GPT-5 as a default model had aroused the dissatisfaction of many users.
With the introduction of GPT-5 last Friday, Openai had withdrawn the possibility of manually selecting its major models of previous language, such as GPT-4.1, GPT-4O, O4-MINI and O3, in order to simplify the use of Chatgpt. The idea being that the system automatically chooses the model most suitable for each task.
However, less than 24 hours after the launch of GPT-5, Openai backtracks by announcing the reintegration of the GPT-4O option-but not for all users. “We will give subscribers plus the possibility of continuing to use GPT-4O”, writes the CEO of Openai, Sam Altman, on X, adding: “We observe the use and assess the duration during which we will continue to offer the old models.”
Openai thus reacts, it seems, to the complaints of many users. Many of them complained about Reddit of the loss of the previous model, including his style of conversation and his tone. As for GPT-5, it is the subject of multiple criticisms: the model would have less “personality” and provide shorter and insufficient responses, according to a very followed reddit thread entitled “GPT5 is horrible”. Another criticism relates to the reduction of the number of permitted queries, which, according to some users, would lead to subscribers more to exhaust their quota in less than an hour.
In his message published on X, Sam Altman promises to double the limits of use of GPT-5 as part of the Chatgpt Plus subscription and this from the end of the current deployment. GPT-5 should also gain “intelligence”. In the future, Chatgpt should also indicate in a transparent manner which model responds to a request.
