Just two days after Announcement GPT-4.1Openai releases not one, but two new models. The company announced today Public availability O3 and O4-MiniThe field from the first, Openai says that O3 is its most advanced model of reasoning, and it shows “strong performance” in the tasks of coding, mathematics and science. As for O4-Mini, Openai issues his account as an alternative to a lower cost, which still gives “impressive results” in the same areas.
Moreover, both models offer new opportunities not found in past Openai systems. For the first time, the companies of the company can use and combine all the tools available in ChatGPT, including Web examination And Generation of imagesThe company says that this possibility allows O3 and O4-Mini to solve complex, multi-stage problems more efficiently and “take real steps to action independently.”
At the same time, O3 and O4-Mini can not only see images, but also interpret and “think” about them in such a way that significantly expands their visual processing capabilities. For example, you can download images of white boards, diagrams or sketches – even low quality – and new models will understand them. They can also configure images as part of how they reason.
“The combined power of modern reasoning with full access to tools leads to significantly higher productivity in academic criteria and real tasks, establishing a new standard both in intelligence and in utility,” says Openai.
Separately, Openai produces a new coding agent (To the code of Claude) named Codex Cli. It is designed to give the developers a minimum interface that they can use to connect the Openai models with the local code. From the box it works with O3 and O4-Mini, with the support of GPT-4.1 in the way.
Today's announcement appeared after the general director of Openai Sam Altman said that the company was changing the course on the roadmap, he Details in FebruaryThe field at that time Altman indicated that Openai would not release O3, which is first the company Previously viewed at the end of last yearas a separate product. However, in early April, he announced the “change of plans”, noting that Openai was moving forward with the release of O3 and O4-Mini.
“There are many reasons for this, but the most exciting is that we can make the GPT-5 much better than we initially,” he wrote on xThe field “It was also more difficult for us than we thought that it would smoothly integrate everything. And we want to make sure that we have enough opportunities to support what we expect to become unprecedented in demand. ”
This means that the optimizing altman promised in February, most likely it will be necessary to wait, at least before the GPT-5 exit, which, according to him, will appear somewhere in the next “several months”.
At the same time, Catgpt Plus, Pro and team users can begin to use O3 and O4-Mini, starting from today. Somewhere in the next few weeks, Openai will bring online O3-Pro, an even more powerful version of its flagship model of reasoning, and will make it available to the PRO subscribers. Currently, these users can continue to use O1-Pro.