OpenAI launches ChatGPT-5, its new generative AI model

OpenAI launched its GPT-5 artificial intelligence model on Thursday, the highly anticipated latest installment of a technology that has helped transform global business and culture. OpenAI's GPT models are the AI technology that powers the popular ChatGPT chatbot, and GPT-5 will be available to ChatGPT's 700 million users, according to OpenAI.

The big question is whether the company that ignited the generative AI frenzy will be able to continue driving significant technological advancements that appeal to business users and justify the huge sums of money it is investing to fuel these developments.

The launch comes at a critical time for the AI industry. The world's largest AI developers—Alphabet, Meta, Amazon, and Microsoft, which backs OpenAI—have dramatically increased capital investment to fund AI data centers, fueling investor hopes for big returns. These four companies expect to invest nearly $400 billion in total this fiscal year. OpenAI is in the early stages of negotiations to allow its employees to earn benefits at a $500 billion valuation, a major step up from its current valuation of $300 billion. Top AI researchers now receive $100 million signing bonuses.

"So far, business spending on AI has been quite low, while consumer spending on AI has been quite strong because people love chatting on ChatGPT," said economist Noah Smith. "But consumer spending on AI won't be enough to justify all the money being invested in AI data centers."

OpenAI highlights GPT-5's business capabilities. In addition to software development, the company said GPT-5 excels at writing, health-related queries, and finance. "I think GPT-5 is really the first time, in one of our core models, that we've felt like you can ask a legitimate expert, a PhD expert, anything," said Sam Altman, OpenAI's CEO, in a press conference.

One of the most exciting things you can do is create quality, instant software. This idea of on-demand software will be one of the defining features of the GPT-5 era.

In demonstrations on Thursday, OpenAI showed how GPT-5 could be used to create fully functioning software based on written text prompts, commonly known as “vibration coding.” A key measure of success is whether the progress from GPT-4 to GPT-5 matches the research lab’s previous improvements. Two early reviewers told Reuters that while the new model impressed them with its ability to code and solve science and math problems, they believe the leap from GPT-4 to GPT-5 was not as great as OpenAI’s previous improvements.

While the improvements are significant, GPT-5 is not advanced enough to completely replace humans. Altman stated that GPT-5 still lacks the ability to learn on its own , a key component for AI to match human capabilities. On his popular podcast about AI, Dwarkesh Patel compared current AI to teaching a child to play the saxophone by reading the notes of the previous student.

"A student only tries once," he said. "As soon as they make a mistake, you fire them and write detailed instructions about what went wrong. The next student reads your notes and tries to imitate Charlie Parker. If they fail, you refine the instructions for the next one. This simply wouldn't work."

Nearly three years ago, ChatGPT introduced the world to generative AI, dazzling users with its ability to write human-like prose and poetry, quickly becoming one of the fastest-growing applications in history. In March 2023, OpenAI followed up on ChatGPT with the release of GPT-4 , a large language model that represented a breakthrough in intelligence. While GPT-3.5, an earlier version, scored in the bottom 10% on the bar exam, GPT-4 passed the mock exam in the top 10%. GPT-4's advancement was driven by increased processing power and data, and the company hoped that similar "scaling up" would result in consistently improved AI models. However, OpenAI encountered challenges scaling. One problem was the data barrier the company encountered, with former OpenAI chief scientist Ilya Sutskever stating last year that while processing power was increasing, the amount of data was not.

He was referring to the fact that large language models are trained on massive datasets that gather information from across the internet, and AI labs have no other options for large amounts of human-generated textual data. In addition to the lack of data, another problem was that training runs for large models are more prone to hardware-induced glitches given the complexity of the system, and researchers might not know the final performance of the models until the end of the run, which can take months.

At the same time, OpenAI discovered another route to smarter AI, called “test-time computing,” a way for the AI model to spend more time computing each question, allowing it to solve complex tasks like math or complex operations that require advanced reasoning and decision-making.

GPT-5 acts as a router, meaning that if a user asks it a particularly difficult problem, it will use time-based computing to answer.

This is the first time the general public will have access to OpenAI's time-tested computing technology, something Altman says is important to the company's mission to develop AI that benefits all of humanity. Altman believes current investment in AI remains insufficient. "We need to build a lot more infrastructure globally to make AI available locally in all of these markets," Altman said.