OpenAI is rolling out a more powerful and long-awaited new artificial intelligence model called GPT-5, vying to stay ahead of increased competition from rivals in the US and China.
GPT-5, set to be unveiled during a live streamed event on Thursday, is designed to be more capable at coding and creative writing as well as reasoning through complex queries. In a briefing with reporters this week, OpenAI Chief Executive Officer Sam Altman called it a “major upgrade” from the company’s prior AI models. For “the first time,” he said, “it really feels like talking to an expert in any topic.”
OpenAI will make GPT-5 widely available to all free users and many of its paid ChatGPT subscribers starting on Thursday, with education and enterprise customers slated to get access the following week. Paid customers will have higher usage limits with the tool.
The San Francisco-based company kicked off the generative AI boom nearly three years ago with the release of ChatGPT, which was originally powered by an earlier model called GPT-3.5. Since then, the company has released a series of increasingly sophisticated systems, including multiple options that mimic the process of human reasoning and AI agents that are meant to field tasks on a user’s behalf with little or no intervention.For much of this year, there has been intense speculation around GPT-5, with some of it fanned by Altman himself. In April, he said GPT-5 would be “much better” than the company originally expected. More recently, Altman expressed some personal apprehension at the new model’s capabilities, relative to his own. “I got emailed a question that I didn’t quite understand and I put it in the model — this is GPT-5 — and it answered it perfectly,” Altman said on a podcast. “I felt useless relative to the AI.”
The new model is quicker to respond to users, better at answering queries and also makes things up less often than prior models, said Nick Turley, who runs OpenAI’s ChatGPT team. “When you’re talking to this thing, it feels just a little bit more natural,” Turley said.As part of the launch, OpenAI is introducing four pre-set personalities that ChatGPT users can choose from to better tailor their interactions with the chatbot. The options, which are initially available as a “research preview,” include Cynic, Robot, Listener and Nerd.
With GPT-5, ChatGPT can also determine when to spend more time ruminating over a user’s query, and how much time to devote to doing so, OpenAI said. Doing this automatically, rather than by making users choose whether to use a typical large language model or a reasoning system, may make the most of the company’s computing resources and give users the most desirable response.
Crucially, OpenAI is positioning the model as particularly helpful for computer programming. A growing number of large tech firms and startups have launched AI assistants intended to help software developers of varying skillsets speed up the process of writing and debugging code, spawning a vibe coding trend. “This idea of software on demand will be a defining part of the new GPT-5 era,” Altman said.
In a live demonstration Wednesday, a researcher showed how GPT-5 could be harnessed to quickly make a web app to help teach French to an English speaker. The researcher asked for a “beautiful, highly interactive” application with activities like flashcards, quizzes and a classic Snake-like game where a mouse eats cheese. Minutes later, ChatGPT spit out the software, including the game.
To test out the model’s capabilities, OpenAI engaged a number of companies as early testers, including Anysphere, maker of the popular AI coding assistant Cursor, and Lovable, a Swedish vibe coding startup.
Lovable co-founder and CEO Anton Osika said in an interview that his company found GPT-5 better than other AI models at quickly building complicated applications, such as a financial planning app that includes a chatbot within it. Software developers typically have to iterate while building such applications, he said, and the new model makes it easier by producing computer code that is easier to maintain. It’s also good for fixing bugs in software, he said.
While OpenAI has long been a frontrunner in the AI market, it faces renewed competition, in some cases from former employees and allies. Alphabet Inc.’s Google, Anthropic and xAI, run by OpenAI co-founder Elon Musk, have released reasoning models and agent products to rival OpenAI’s offerings. They’re also squarely focused on the rapidly expanding market for AI-assisted coding. Chinese firms like DeepSeek are gaining ground, too. And Meta Platforms Inc. is spending heavily to hire top talent for a new AI lab, including poaching more than a dozen employees from OpenAI.
In hopes of keeping an edge, OpenAI, now valued at $300 billion, has raised tens of billions of dollars to help cover the cost of talent, chips and data centers needed to build and support cutting-edge models, including GPT-5. OpenAI is also in early talks about a potential sale of stock for current and former employees at a valuation of about $500 billion, Bloomberg News has reported.
OpenAI’s not-so-secret weapon continues to be ChatGPT, the chatbot that made the company a household name, which serves as a gateway product for its various paid services. OpenAI said it now has nearly 700 million ChatGPT users each week and 5 million paying business users.
On Wednesday, OpenAI said it would provide access to ChatGPT to US federal agencies at a nominal cost of $1 a year as part of a push to get its chatbot more widely adopted. OpenAI also released a pair of free, open models earlier in the week to rival offerings from DeepSeek and Meta.
