OpenAI announced the launch of its newest and most advanced artificial intelligence model, Chat GPT-5. According to the company, the version is smarter, faster, and more widely useful than previous ones, with dramatic upgrades in writing capabilities, computer programming, and medical applications.
Sam Altman, CEO of OpenAI, explained during the briefing that when he tried to go back to GPT-4, the experience was "shocking," he said, due to the significant upgrade felt on every level. According to him, using GPT-5 feels like "working with a team of doctors in various fields, at any given moment."
Altman noted that the company has reached a peak of 700 million active users per week on Chat GPT, an amazing figure that reflects the wide penetration of the tool into everyday life. OpenAI is currently also in talks with investors for an IPO at a valuation of about half a trillion dollars.
The new version presents a lower rate of "hallucinations," meaning invention of incorrect facts, a known problem in artificial intelligence. According to the company, GPT-5 underwent a strict safety process including over 5,000 hours of testing and meticulous examination. Unlike previous versions, GPT-5 does not automatically refuse to answer problematic questions but offers general and safe answers that cannot be misused.
Michelle Pokras, one of the heads of the model development team, noted that GPT-5 learns to recognize when a task cannot be completed, avoid unnecessary guesses, and explain limitations more clearly. This ability reduces the number of false claims compared to previous models.
During the demonstration, a new concept called "vibe coding" was presented: Writing complete software based on simple instructions. For example, the engine was asked to build an app for learning French for English-speaking users, including flashcards, quizzes, and daily tracking. Within seconds, GPT-5 created two completely different apps, without requiring prior coding knowledge. Each app had a different interface and the option to continue improving it.
At the same time, OpenAI also announced three additional versions of GPT-5 for developers: Pro version, Mini version, and Nano version, allowing adaptation between cost, performance, and response speed. The company released two open models, the first time since GPT-2 in 2019, for developers interested in running and customizing the model themselves.
According to the company, the conversational interface with GPT-5 feels more natural, with responses that sound almost human in structure and nuances. For free users, this step will enable access to a model based on internal reasoning, meaning one that runs a chain of thought before providing an answer. If users exceed the usage quota, they will get access to the GPT-5 Mini version. Plus users enjoy a wider usage quota, Pro users receive unlimited access and the option to run the GPT-5 Pro version, and ChatGPT Enterprise and Edu users will receive the upgrade too.