If the speculation about GPT-5 is not new, the CEO of Openai, Sam Altman, said that GPT-5 would soon be available on X at the start of the week, new clues suggest that it may be earlier than expected. According to The Verge, which is based on internal sources, the new GPT-5 model could be available in early August and it will be associated with mini and nano versions, which will also be made available to developers via an API.
(Information: Ziff Davis, the parent company of ZDNET, brought legal action against OpenAI in April 2025, accusing him of having violated his copyright during the training and exploitation of his AI systems.)
Here is everything you need to know, updated with the latest information available.
A brief overview of the chronology
When Chatgpt experienced its peak in November 2022, it operated on GPT-3.5. Openai then published GPT-4 in March 2023, which, at the time, marked a major advance in intelligence. Then an improved version, GPT-4O, arrived in March 2024 introducing multimodality. Today, a little more than a year later, GPT-5 is late, which was not planned at the start.
In February, Sam Altman shared a roadmap for X concerning new corporate models, specifying in the comments that it would only take a few weeks/months. However, in April, the big boss announced a change of plan, switching to the exit of O3 and O4-Mini and pushing GPT-5 a few months. Sam Altman justified this delay by the difficulties encountered during the integration of the various elements into the model.
What do we know about GPT-5
GPT-5 combines the reasoning capacities of models in the O series and models in the GPT series. The system would be able to understand when to make the best use of the assets of the two types of models, which represents a considerable advantage for users, as this guarantees them the best combination of speed, cost and quality.
Although the reasoning models are effective in producing quality responses to complex problems, their implementation is longer because they require in -depth reflection, are therefore more expensive and are not always necessary if the task does not require it. On the other hand, if Chatgpt or the user systematically opts for default GPT models, reasoning models lose opportunities to produce better answers. Currently, users have the opportunity to switch between the two, but often they do not know exactly which model is the most suitable.
Sam Altman said that free chatgpt users will benefit from unlimited access to the GPT-5 cat with the standard intelligence level, while subscribers more can access a higher level of intelligence. Pro subscribers will be able to use it at the highest level. The models should combine all the best features of Chatgpt, including voice, canvas, research and in -depth research.
The CEO of Openai recalled last week that the model will be experimental and will integrate different research techniques, but that it is not the anonymous model rewarded by a gold medal during the last International Mathematics Olympiad (IMO).
