- Chatgpt users paid are receiving access to a lot of older models
- Altman mocks an update to the personality of ChatgPT-5
- The new “warmer” update is to counteract the reaction to GPT-5
In his last tweet on the Social Network Platform X, Sam Altman, CEO of OpenAI, has confirmed that all paid Chatgpt subscribers will get access not only to the old GPT-4o model, but also LLMS as O3, 4.1.
The popular Chatgpt-4.5 will also return, but it will only be available for professional subscribers. Altman says that this is because “it costs many GPUs”, a reference to the amount of calculation power it requires.
Following the violent reaction against the elimination of the popular 4th model without absolutely no warning when GPT-5 was launched, Altman seems to have learned a lesson and has promised: “If we ever discourage it, we will give close attention.”
All paid chatgpt users should now find an alternation of ‘show additional models’ in the chatgpt web configuration, which will give access to all older LLM models. You can also add a new GPT-5 Thinking Model.
Chatgpt updates: You can now choose between “Auto”, “Fast” and “Think” for GPT-5. Most users will automatically want, but additional control will be useful for some people. The rates limits are now 3,000 messages/week with GPT-5 thought, and then additional capacity in GPT-5 thinking …August 13, 2025
Altman also refers to the highly criticized ‘colder’ tone of the new Chatgpt-5, which has moved many users on the tweet: “We are working on an update of the GPT-5 personality that should feel warmer than the current personality, but not so annoying (for most users) as GPT-4O”.
His reference to ChatgPT-4O is annoying refers to the Sycophic phase in which GPT-4O seemed to enter after an update in April.
Altman continues: “However, an learning for us in recent days is that we really just need to reach a world with more customization by user of the model personality.”
Multiple personalities
Altman’s reference to “user customization” reflects OpenAi’s recognition that what its users want is an easier way to select how formal, humorous, empathetic or direct is the assistant.
If they gave me free rays to imagine how I would like the chatgpt to work, I would like to reach the stage where Chatgpt personality traits could be represented through sliding controls, such as ‘Professional versus Casual’ or ‘Concise versus detailed’. That would make it much easier to get the results you are looking for.
While customgpts already exists, I would love it to be easily changing between personality types, such as ‘work assistant’ or ‘creative writing coach’. However, I have the feeling that it will be a long time before we have a chatbot of the so easily customizable to speak.
You may also like
