OpenAI has announced two new AI models that focus on improving reasoning. The OpenAI o3 model is described by its developers as “the most powerful reasoning model ever.” And the OpenAI o4-mini is a smaller, faster reasoning model that delivers “impressive performance for its size and cost.”
Image source: OpenAI
A special feature of the new large language models is their ability to “think” in images, that is, to integrate visual information directly into the chain of reasoning. This is especially useful when working with sketches or content on electronic boards. The models can also modify images – enlarge, rotate, and analyze them as they are processed.
OpenAI also says the new models will be able to use all of ChatGPT’s tools, including web search, image analysis and generation, and file reading. These features are available starting today to ChatGPT Plus, Pro, and Team customers using the o3, o4-mini, and o4-mini-high models. Support for the most powerful o3-pro model is expected “in the coming weeks.” The current o1, o3-mini, and o3-mini-high models will be phased out on those plans.
Today’s announcements follow the unveiling of the company’s flagship AI model GPT-4.1, the successor to GPT-4o, on Monday.