OpenAI launches o3 and o4-mini, AI models that ‘think with images’ and use tools autonomously [English]

By Business Telegraph | Read the full article in English

Breakthrough in AI Technology | Visual Reasoning Capabilities | Advanced Problem Solving

OpenAI has just unveiled two groundbreaking artificial intelligence models that represent a significant leap forward in machine learning technology. The new o3 and o4-mini models can do something truly remarkable: they don't just look at images, they actually think about them. This means the AI can now analyze visual information, break down complex diagrams, and solve problems in ways that were previously impossible for computer systems.

These new models are incredibly versatile, capable of performing complex tasks that would typically require human intervention. They can search the internet, run computer code, analyze files, and even generate images as part of solving a problem. For example, if someone asks the AI to research future energy patterns in California, it can independently gather data, write analysis code, create visualizations, and produce a comprehensive report – all without constant human guidance.

The technological advancement goes beyond just processing information faster. OpenAI claims these models set new benchmarks in AI capabilities, with the smaller o4-mini model scoring an impressive 99.5 percent on a mathematics competition. The company has also emphasized extensive safety testing to ensure the models can refuse potentially harmful requests. This release signals a significant moment in AI development, where machines are beginning to perceive and interact with visual information in a way that more closely mimics human cognitive processes.

Read More (translated)