OpenAI unveiled o3-pro — an updated version of its AI model for logical reasoning.

OpenAI has released o3-pro, an enhanced version of its reasoning model o3, which is considered the most powerful to date.

Unlike traditional AI models, reasoning models solve tasks step by step, making them more reliable in fields such as physics, mathematics, and programming.

O3-pro is available to ChatGPT Pro and Team users starting June 10, 2025, replacing the previous o1-pro version. Access for enterprise clients and educational institutions will be available next week. The model is also already integrated into the API for developers.

The cost of using o3-pro in the API is $20 per million input tokens and $80 per million output tokens. According to OpenAI's estimates, one million input tokens is equivalent to approximately 750,000 words — slightly more than the novel "War and Peace."

Internal testing showed that o3-pro outperforms o3 across all key metrics, including science, education, programming, business, and writing assistance. The model also received high marks for clarity, completeness of responses, and accuracy in following instructions.

O3-pro is equipped with tools for internet search, file analysis, visual information processing, Python execution, and personalized responses using memory. However, due to this, response times may be longer than those of o1-pro.

In OpenAI's tests, o3-pro surpassed Google Gemini 2.5 Pro in mathematical tasks and outperformed Anthropic's Claude 4 Opus in a PhD-level knowledge test.

At this time, the model does not support image generation, and the instant chat feature in ChatGPT has been temporarily disabled due to technical settings.

Source: TechCrunch