OpenAI's long-awaited GPT-5 model is nearing release

OpenAI's GPT-5, the latest version of the AI technology that underpinned the ChatGPT giant in 2022, is expected to be released soon, and users will be carefully checking whether this step compared to GPT-4 is equal to previous achievements of the research laboratory, Reuters reports, writes UNN.

Details

Two early testers of the new model told Reuters they were impressed by its coding capabilities and ability to solve scientific and mathematical problems, but felt that the leap from GPT-4 to GPT-5 was not as significant as the leap from GPT-3 to GPT-4.

The GPT-4 leap was based on increased computing power and data volume, and the company hoped that similar scaling would consistently lead to improved AI models, the publication writes.

However, OpenAI, backed by Microsoft and now valued at $300 billion, has faced scaling challenges. One of the problems was the data wall the company encountered. Former OpenAI chief scientist Ilya Sutskever said last year that despite the growth in computing power, the volume of data did not grow.

He was referring to the fact that large language models are trained on vast datasets that scan the entire internet, and AI labs have no other options for working with large amounts of human-generated text data.

In addition to the lack of data, another problem was that "training runs" of large models were more likely to have hardware failures, given the complexity of the system, and researchers might not know the final performance of the models until the end of the run, which could take several months.

OpenAI has not announced a release date for GPT-5, but media reports suggest the industry is expecting it any day now. Boris Power, head of applied research at OpenAI, wrote in his X post on Monday: "Looking forward to seeing how the public receives GPT-5."

"OpenAI made such a huge leap from GPT-3 to GPT-4 that GPT-5 has generated huge anticipation ever since," said Naveen Chaddha, managing partner at Mayfield, a venture fund that invests in AI companies but is not an OpenAI investor. "We hope GPT-5 will unlock access to AI-powered applications that go beyond chat and allow for fully autonomous task execution."

Addition

Almost three years ago, ChatGPT introduced the world to generative AI, impressing users with its ability to write human-like prose and poetry, and quickly becoming one of the fastest-growing applications in history.

In March 2023, OpenAI followed ChatGPT with the release of GPT-4, a large language model that made a huge leap forward in intelligence. While GPT-3.5, an earlier version of the model, scored in the bottom 10% on the bar exam, GPT-4 passed a simulated bar exam in the top 10%.

GPT-4 became a benchmark model, and the world recognized that AI models could outperform humans in many tasks, the publication notes.

Soon, other companies adopted this approach. In the same year, Google (Alphabet) and Anthropic, backed by Amazon and Google, released models that competed with GPT-4. Within a year, open-source models that were comparable to GPT-4 were released, such as Meta Platforms' Llama 3 models.

In addition to training large models, OpenAI is investing in another direction called "test-time compute," which allows for the use of significant computing power to solve complex tasks, such as mathematical calculations or complex operations that require human thinking and decision-making.

Earlier this year, CEO Sam Altman stated that GPT-5 would combine both test-time compute and large models. He also noted that OpenAI's models and products have become "complex."

OpenAI hits $12 billion in annualized revenue - report31.07.25, 11:14 • 3826 views