ChatGPT-5 is supposed to be „warmer than the current personality but not as annoying,” and 4o is staying with us for now, Sam Altman promises after failed launch

OpenAI promised that ChatGPT-5 will be a real revolution among artificial intelligence models. These words were also spoken during the presentation of the new version of the chatbot, when Sam Altman referred to it as a "conversation with a doctor of science" instead of a student or pupil, which the older versions were supposed to represent. However, this promise was not fully realized, which was evident in the users' reactions. Because of that, the AI model will receive changes.

Limits for ChatGPT-5 Thinking

On his X profile, Sam Altman outlined what changes were introduced in the ChatGPT-5 update. First, there are three versions of the model: Auto, Fast, and Thinking. The default is Auto, but the CEO of OpenAI has indicated that the other two may be useful for people requiring more control. Besides that, he also presented new limits in the operation of the chatbot:

1. 3000 messages weekly with GPT-5 Thinking;
2. Additional bandwidth for GPT-5 Thinking mini after exceeding the limit;
3. The context limit for GPT-5 Thinking is 196000 tokens.

The values of these limits are not final and may still change according to the head of OpenAI.

ChatGPT-5 will be more human-like in conversations

OpenAI is also working on the personality of the new model. It can become "gentler" (some compared ChatGPT-5 to a corporate boss), but not as annoying as the 4o. The behavior of the previous model is a point of contention among the community, because some did not like its overly enthusiastic tone. Now OpenAI is looking for a balance that will be a compromise for everyone.

For others, however, access to ChatGPT-4o may be a more important issue. Initially, the 4o model disappeared with the release of the new version, but this was met with such an intense reaction from the community that it eventually returned. To clarify: ChatGPT-4o is available for subscribers, and besides it, you can also choose GPT-3, GPT-4.1, and GPT-5 Thinking mini models. Altman added that only those with a Pro subscription will benefit from ChatGPT-4.5, as its operation consumes too much GPU computing power.

Sam Altman also promised that if there are plans to abandon the 4o model in the future, the company will provide appropriate information in advance.

• tech
• AI / Artificial Intelligence
• chatbots / ChatGPT