OpenAI announced on the 16th (local time) the release of a new artificial intelligence (AI) model 'o3' capable of solving complex problems and inferring visual information, along with a small model 'o4 mini'.
This model is a successor to the reasoning model 'o1' released last September, equipped with the ability to analyze and understand images such as pictures and charts. OpenAI noted that 'o3' exhibited superior performance compared to previous models by comprehensively testing mathematical, coding, reasoning, scientific, and visual understanding abilities.
Users can input various images, such as whiteboards, drawings, and charts, into 'o3', allowing the AI to interpret them and conduct reasoning. It can recognize blurry or low-quality images and also supports zooming and rotating tasks. Additionally, it is possible to execute Python code directly in the browser or conduct web searches.
These models are available to paid users of ChatGPT Plus, Pro, Teams, etc., and an 'o4 mini high' version with improved response reliability is also provided. In the future, 'o3' will have a higher model 'o3-Pro' released for ChatGPT Pro subscribers.
Sam Altman, OpenAI Chief Executive Officer (CEO), said, 'o3 and o4 mini could be the last standalone AI reasoning models before the launch of GPT-5.'