Sam Altman introduced GPT-5, the new version of ChatGPT, but experts believe it's far from what was promised
In internal tests, GPT-5 achieved a 74.9% accuracy rate on the SWE-bench Verified benchmark, surpassing competitors such as Claude Opus 4.1 and Gemini 2.5 Pro.
It can create websites from scratch, design applications, and explain its process step by step.
Sam Altman took the stage to showcase GPT-5, the latest version of ChatGPT that promises to be faster and smarter.
However, specialists point out that the new Artificial Intelligence is far from what the public expected and doesn't significantly surpass its predecessors.
The CEO of OpenAI caused high expectations with cryptic messages and a major display to present GPT-5.
He promised a "superhuman AI" capable of outperforming the best programmers, but the final product disappointed several experts.
Matt Shumer, founder of OthersideAI, acknowledges that GPT-5 is OpenAI's most advanced model to date.
But he points out that for the average user, the improvements are barely noticeable compared to GPT-4.5, except in programming and code debugging.
GPT-5 features an aesthetic redesign influenced by Jony Ive, with new personalities and customization options available only to paid users.
However, it doesn't break the ceiling in areas where previous versions still dominate, such as scientific research and creative writing.
The development faced issues with a lack of clean data and technical failures during training.
Altman admitted that the launch was delayed to ensure stability and to meet record demand.
