Meet OpenAI’s new AI models o3 and o4-mini, that can think deeper, analyse images, and use tools

OpenAI

OpenAI has just introduced two new AI models — o3 and o4-mini — and they’re the smartest ones the company has built so far. These models are designed to think more carefully before responding, which means they can handle trickier, multi-part questions much better than before.

So what’s new? Well, for the first time, these models can use all the tools inside ChatGPT  like searching the web, running code, looking at images, analyzing files, and even generating pictures. But the real deal is how they decide when and how to use these tools and quickly come up with thoughtful answers in the right format.

Let’s break it down:

–o3 is the bigger, brainier model. It’s great at tough stuff like coding, math, science, and breaking down complicated problems. It’s especially sharp with images, charts, and graphics — and in expert tests, it made 20% fewer big mistakes than earlier versions.

–o4-mini is the smaller, faster, and more budget-friendly option. It’s surprisingly good at math, coding, and visual tasks for its size. Plus, it can handle more queries at once, making it perfect for high-volume use.

Both models are better at following instructions, giving more useful answers, and sounding more natural in conversations. They also remember details from past chats to keep replies relevant and personal.

A cool new feature is their ability to “think” with images. You can upload a photo of a sketch, chart, or even a blurry whiteboard, and they can understand and work with it rotating, zooming, or editing images while reasoning through problems.

Overall, these updates mark a big step forward for AI smarter, faster, and more helpful tools that can tackle real-world problems in clever new ways.