"A team of experts with doctoral degrees": OpenAI showed a new AI model of GPT-5

OpenAI, which created the ChatGPT chatbot, released the new GPT-5 artificial intelligence model. It is already available to all categories of users. The new model is smarter, faster and "much more useful," especially in areas such as writing, programming and healthcare, the developers claim. At the presentation, OpenAI CEO Sam Altman called the new model an important step toward general artificial intelligence (AGI) that will equal or surpass human intelligence.

OpenAI on August 7 launched its new GPT-5 AI model: on Thursday, it became available in a free version, as well as to users with Plus, Pro and Team subscriptions. This is the first time the company has made access to a model with a full-fledged reasoning engine free, notes CNBC. Customers with educational and enterprise plans (ChatGPT Edu and Enterprise) will have access to GPT-5 in about a week.

"GPT-5 is a major upgrade over GPT-4 and an important step towards general artificial intelligence (AGI)," said OpenAI CEO Sam Altman in a video broadcast on YouTube. - We think you'll enjoy using GPT-5 much more than any other AI."

Altman explained that he tried switching to the GPT-4 version after testing the new AI, "and it was pretty dismal." He characterized the new model as "a team of experts with PhDs that are always there for you." The company assured that GPT-5 is much less likely to produce fictitious answers and non-existent facts.

"GPT-5 is trained to recognize tasks it can't complete, avoid speculation and better explain its own limitations," quoted CNBC's Michelle Pokrass, head of post-learning at OpenAI. - This helps reduce the number of unsubstantiated claims compared to previous versions."

Box, an enterprise data management company, has been testing GPT-5 on different types of tasks for the past few weeks. The company noted that previous AI models failed at the most complex tasks, especially those that required understanding math or logic in long documents. But GPT-5, they said, was "a real breakthrough.

"This model is able to hold much more information in memory and use much more powerful logic and reasoning abilities to make decisions," Box CEO Aaron Levie told CNBC.

OpenAI first launched ChatGPT in 2022 and has remained a leader in the artificial intelligence field ever since. The company released its first "reasoning" model o1 in September 2024, and then unveiled o3 in April 2025. OpenAI's product is not unique: Google showed its reasoning-capable Gemini model in December 2024, and Anthropic and Ilon Musk's startup xAI unveiled their "reasoning" models in February, reminds Barron's.

This is OpenAI's second attempt to release a GPT-5. In February, the company presented the GPT 4.5 model, which was designed to be the fifth version, but fell short of its goals in a number of ways.

OpenAI reported that the number of weekly active ChatGPT users grew from 300 million in December to 500 million in March. The AI startup said Thursday that the number of weekly active ChatGPT users will reach 700 million this week - four times more than last year.

OpenAI is now in talks with investors to sell some shares at a valuation of about $500 billion, Bloomberg and the Financial Times wrote, citing sources familiar with the talks.

This article was AI-translated and verified by a human editor