OpenAI Unveils GPT-5: A Smarter, Faster AI Model for Wide-ranging Applications, Including Writing, Coding, and Healthcare
OpenAI unveils its most advanced AI model yet, GPT-5, offering enhanced capabilities across various domains including writing, coding, and healthcare. The company is making the model accessible to all users, including free account holders.
In a briefing with media outlets, OpenAI CEO Sam Altman highlighted GPT-5’s improvements over its predecessor, GPT-4, noting significantly increased intelligence, speed, and usefulness. Altman described returning to GPT-4 as an unpleasant experience in contrast.
Since the launch of OpenAI’s AI chatbot ChatGPT in 2022, the company has gained significant traction, with expectations to reach over 700 million weekly active users on ChatGPT this week. OpenAI is reportedly in discussions with investors regarding a potential stock sale at a valuation of approximately $500 billion.
GPT-5 boasts a lower hallucination rate, meaning it fabricates answers less frequently. The model underwent extensive safety evaluations during development, including 5,000 hours of testing. Instead of refusing to answer potentially risky questions, GPT-5 will employ “safe completions,” providing high-level responses within safety constraints that cannot be used to cause harm.
Michelle Pokrass, a post-training lead at OpenAI, stated, “GPT-5 has been trained to recognize when a task cannot be completed, avoid speculation, and clarify limitations more clearly, reducing unsupported claims compared to prior models.”
During the briefing, OpenAI demonstrated GPT-5’s capabilities in “vibe coding,” where software is generated with AI based on a simple written prompt. For instance, GPT-5 was asked to create a web app for an English speaker learning French, incorporating engaging themes and activities like flashcards, quizzes, and progress tracking. The model produced two unique apps within seconds, albeit with some rough edges that users can refine as needed.
GPT-5 will be rolled out to OpenAI’s Free, Plus, Pro, and Team users on Thursday. This marks the first time that free users have access to a reasoning model. For users who exceed their usage limits, a scaled-down version called GPT-5 mini will be available.
OpenAI’s Plus users benefit from higher usage limits, while Pro users gain unlimited access to GPT-5 and its advanced features. ChatGPT Edu and ChatGPT Enterprise users can expect access to GPT-5 approximately a week following the initial release.
Box, a company specializing in managing enterprise computer files, has been testing GPT-5 across various data sets recently. Aaron Levie, CEO of Box, expressed that previous AI models failed many of the company’s advanced tests due to their struggles with understanding complex math or logic within lengthy documents. However, he described GPT-5 as a “complete breakthrough.”
“The model is able to retain way more of the information it’s processing and uses a much higher level of reasoning and logic capabilities to make decisions,” Levie told CNBC in an interview.
OpenAI will release three different versions of the model through its application programming interface (API) for developers. These versions – gpt-5, gpt-5-mini, and gpt-5-nano – cater to various cost and latency requirements.
Earlier this week, OpenAI released two open-weight language models as low-cost options since the rollout of GPT-2 in 2019. These models are designed for easy usage and customization by developers, researchers, and companies.
However, with GPT-5, OpenAI aims to cater to a broader consumer audience. The company emphasizes that interacting with the model feels natural and more human-like. Altman described using GPT-5 as having Ph.D.-level experts at one’s disposal anytime.
“People are limited by ideas, but not really the ability to execute, in many new ways,” he said.