Openai unleashes GPT-5-But can he hold out in the ruthless AI race?

One of the decisive truths about the generating AI world is that even when you are at the top, the advance does not last long.

And so, the two key questions coming out of the long-awaited launch of GPT-5 OpenAi today are whether the new LLM can help the company recover the indispensable IA leader’s coat-and if so, how long can the head keep?

OPENAI says that the GPT-5 offers “more precise responses than any model of previous reasoning” and is “much more intelligent in all areas”, reflected by strong performance on academic references and evaluated by humans. His research blog has new advanced performance through mathematics, coding and health issues, and found that GPT-5 has surpassed other OPENAI models through tasks covering more than 40 professions, including law, logistics, sales and engineering.

“GPT-5 really wants to talk to a doctoral level expert in any subject,” the CEO of Openai, Sam Altman, in a prerequisite on Wednesday. “Something like GPT-5 would be almost unimaginable at any other moment in history.”

Altman described the GPT-5 as an “important step” on the path of artificial general intelligence (AGA), which, according to OpenAi’s mission statement, is defined as “highly autonomous systems that surpass humans to the most economically precious work”.

It is not clear if this combination of speed, power and characteristics will be sufficient, however. Some two years in manufacturing (GPT-4 was launched in March 2023), the launch of the GPT-5 took more time than many initiates in the industry did not plan, because Openai adjusted its approach in response to industry changes. And although Chatgpt now has an impressive 700 million weekly users, Openai has faced increasing pressure in the past year while its rivals have poached its talent and continue on emerging AI techniques such as long context and the use of autonomous tools. In addition to major technological competitors like Meta and Google, there is a wave of startups founded by former Openai researchers, including anthropogenic machines, thinking machines and a sure superintelligence. And of course, there is the new harvest of powerful Chinese models, like Deepseek, in the running for global influence.

The fact that GPT-5 propels Openai at the top of the Hill AI will become clear in the days and weeks in advance, because the researchers put the model to the test, testing it against other elite models, including the latest Claude d’Anthropic model and the Gemini de Google.

Openai pushes to stay in the lead

With GPT-5 now, the CEO of Openai, Sam Altman, recognized that staying on the border means one thing: an implacable scaling.

In AI, scaling refers to the idea that models become more powerful as you increase the amount of data, calculation power and model components used during training. This is the underlying principle that has led to progress from GPT-2 to GPT-3 to GPT-4-and now GPT-5. The capture is that each jump requires investment exponentially more

When asked if the scaling laws were still held, Altman said they were doing it “absolutely”. He stressed that better models, smarter architectures, better quality data and much more computing power, because the path of “order of grandeur” improvements is still to come.

But this kind of progress has a cost. “This will take a control amount of calculation,” he admitted. “But we intend to continue to do so.”

This, of course, requires a sum of money and massive partnerships.

On the right side, Openai approached its income in the first seven months of 2025, reaching an annualized execution rate of $ 12 billion, against around $ 6 billion at the start of the year, according to a recent information report. This translates into $ 1 billion in monthly income, fueled by demand for demand for its chatgpt products on consumer and companies markets. Weekly chatgpt active users jumped at around 700 million, compared to 500 million from all OPENAI products at the end of March. And earlier this week, Openai published a free and open source model-an unusual decision for an company often criticized for its closed approach during the last half-receiver, which maintains that its premium offer, which is now GPT-5, will continue to dominate.

However, there are great challenges to come. On the one hand, the partnership between Microsoft and Openai – which started with an investment of $ 1 billion in 2019 – entered a heavier and more complex phase. While Microsoft has invested more than $ 13 billion and retains exclusive rights to OPENAI models via Azure, tensions have appeared on income sharing, AG control clauses and overlapping products. And the Stargate project depends on partnerships with companies like SoftBank, whose investment stipulations are linked to the always unrexoid efforts of Openai to revise its corporate structure.

There is much more to say about GPT-5. Journalists have received the full whole of OpenAi documents, including a search blog, a system card and a security card, barely 90 minutes before publishing the model, so I still have a lot to cross. Stay listening to find out more!

Also: in less than a month, I will head for Park City, Utah, to participate in our annual conference Brainstorm Tech in the Deer Valley assembly! The space is limited, so if you are interested in joining, register here. I highly recommend: there is a fantastic range of speakers, including Ashley Kramer, Openai income chief; John Furner, president and chief executive officer of Walmart US; Tony XU, founder and CEO of Doordash; And a lot, much more!

With that, here are the rest of the news from AI.

Sharon Goldman
sharon.goldman@fortune.com
@Sharongoldman

Have in the news

OPENAI concludes a government agreement at $ 1. OPENAI has concluded an agreement with the US government to provide federal agencies with access to its border AI models – including Chatgpt – for only $ 1 in next year, according to a joint announcement of the General Services Administration (GSA). The partnership reflects months of awareness behind the scenes by the CEO of Openai, Sam Altman and his team, which have cultivated links with the Trump administration since before Donald Trump’s return to the White House in January. In a declaration at CableAltman supervised the agreement as part of President Trump’s Action Plan, saying that it would help civil servants take advantage of AI to better serve the American people.

AID AIDS OF AI are detached. A week after Openai published a Chatgpt study mode, Google jumped into the AI tutor game with a new guided learning mode. Google says that this has worked in close collaboration with learning experts, teachers and students to develop guided learning, which is integrated into Gemini. Like Chatgpt study mode, Google’s guided learning is designed to encourage users to solve problems with “survey and open questions”, rather than simply spitting the answers. With courses in many schools and universities that should start in the coming weeks, the big question will be whether the students really want an AI study companion or are content with an AI machine.

Zoox obtains an exemption from the steering wheel. The evolution of Robotaxi took a step forward Wednesday while the American federal regulators authorized the Amazon zoox to test the autonomous shuttles without steering wheel, pedals or other manual controls. This decision clarifies an ambiguity point in the autonomous automotive industry. While Zoox had previously said that its vehicles complied with the federal standards of motor vehicles, the rules of the National Highway Traffic Safety Administration seemed to suggest that this was not the case. The NHTSA decision clearly indicates that autonomous cars built in the United States are indeed needing manual controls to operate (even for tests) on public roads, but grant Zoox an exemption – and opens the door to other Robotaxi companies such as Tesla to obtain a derogation.

Fortune we have

The AI has already turned the picture of corporate organizations upset because it flattens the distance between the suite C and everyone – by Béatrice Nolan

How Palant – A company too small to make fortune 500 – has become one of the 25 most precious companies in the world – by Jessica Matthews

OPENAI is launching its first model open for years so that it can stop being “bad side of history” – while always keeping its most precious IP under Wraps – by Sharon Goldman

You have a calendar

September 8-10: Fortune Brainstorm Tech, Park City, Utah. Apply to attend.

October 6-10: World IA Week, Amsterdam

October 21-22: Tedai San Francisco. Apply to attend.

December 2-7: Neirips, San Diego

December 8-9: Fortune IA San Francisco. Apply to attend.

Eye on AI numbers

$ 364 billion

The total amount of money that the four highest tech – Meta, Microsoft, Google and Amazon Hyperscales – plan to spend 2,25 capital expenses when they build their IA and Cloud infrastructure. Companies pay enormous sums of money in the construction and expansion of costly GPU NVIDIA data centers in a race to stay ahead in AI and to reach AG.