OpenAI Launches New AI Models Capable of Image and Text Reasoning

On April 16, 2025, OpenAI introduced two advanced artificial intelligence systems, OpenAI o3 and o4-mini, designed to reason through tasks involving both images and text. These new models expand on capabilities introduced in earlier versions by being able to manipulate, analyze, and interpret images such as diagrams, sketches, and graphs in addition to processing language. According to OpenAI’s head of research, Mark Chen, the systems can perform complex operations like transforming and cropping images to support various tasks. One notable improvement is the models’ ability to ‘think’ before responding, mimicking human-like step-by-step reasoning.

OpenAI also launched Codex CLI, a new AI agent intended to support programmers by integrating with existing code on their local machines. This tool is open-sourced, allowing developers and businesses to adapt and build on it freely.

These innovations reflect a broader industry trend toward creating AI that can reason through complex challenges. While promising, experts caution that such systems do not reason like humans and may still produce incorrect or fabricated information, a phenomenon known as ‘hallucination.’ Access to o3 and o4-mini is available to ChatGPT Plus and Pro subscribers, costing $20 and $200 per month, respectively.