OpenAI unveils GPT-5 today — key details inside

OpenAI will be hosting a livestream today, August 7, at 10AM PT, where it will likely officially showcase GPT-5 for the first time. It is hinted at by the stylized number "5" in the event title and new model icons spotted in the company's API.

The company announced it on its X page.

What to expect from GPT-5

OpenAI is preparing three variations at once — GPT-5, GPT-5-mini, and GPT-5-nano. The most compact nano version will work only through the API and is focused on fast and inexpensive answers. The basic GPT-5 will receive several modes with different "depth of thought": the longest chains of reasoning will become available only to subscribers of the Plus and Pro tariffs.

The main idea of the new model is "versatility": it will independently determine when to give an instant result and when to proceed to a detailed logical analysis. For enthusiasts, the option "to think longer" will remain.

An important update will also affect accuracy. During reinforcement training, OpenAI used the Universal Verifier module — an auxiliary AI that checks GPT-5 answers and sends doubtful ones for revision. Thanks to this, the developers promise a significant reduction in the number of hallucinations not only in programming or mathematics tasks, but also in creative texts and business scenarios.

Read also:

ChatGPT's new study mode could replace your teacher

ChatGPT vs Google — AI gets 2.5B daily queries