OpenAI announced one of the most ambitious artificial intelligence updates of this year. The company’s new model O3-pro stands out as an advanced and expanded version of the previously introduced O3 artificial intelligence engine. The OpenAI O3-pro attracts attention with its more consistent and solid results, especially in problem solving and logical reasoning processes.
The main difference of the model is that it has the ability to think step by step by separating from classic artificial intelligence systems. This feature makes the model more reliable, especially in areas such as mathematics, physics, software and technical analysis. O3-pro facilitates users to produce more configured solutions to complex problems.
OpenAI promises more accurate and clear answers with O3-pro
The new model has been activated for Chatgpt Pro and Team users as of June 11th. OpenAI, corporate (enterprise) and education (EDU) users will start to use the model within a week, he said. On the other hand, O3-pro for developers, OpenAI API on the same day was opened to access.
O3-Pro’s API usage costs were also announced. 1 million input (input) token is 20 dollars and 1 million output (Output) token is requested 80 dollars. The entrance represents the data presented to the Model by the user and the exit token represents the answer produced by the model. This data volume, which corresponds to about 750,000 words, makes it possible to work with novel length content.
OpenAI also shared the comprehensive internal test results of the model with users. Experts, O3-pro’nun predecessor O3 compared to each test category is more successful results, he says. In these evaluations, it is emphasized that it stands out in areas such as science, software, education, business correspondence and text arrangement. In addition, it is stated that the model performs higher in criteria such as clarity, details, harmony with instructions and accuracy.
O3-Pro’s technical capabilities include file analysis, use of Python, internet screening and logic on visual content. The model can also offer personalized responses based on the user’s past preferences. Nevertheless, O3-pro’s response production time is said to be a little slower than the previous model O1-pro.
In addition, some limitations of O3-pro may also affect the user experience. The temporary chat feature on the chatgpt platform is currently disabled due to a temporary technical problem. O3-pro does not have visual production feature, and OpenAI’s AI-supported work area is not compatible with this model for now.
However, the achievements of the model in various artificial intelligence tests put the current limitations to the second place. According to OpenAI, O3-pro, the AIME 2024 test, which measures mathematical competence, left Google’s leading model Gemini 2.5 Pro. However, the GPQA Diamond test, which evaluates the knowledge of science at the level of doctoral level, also performed better than Anthropic’s Claude 4 Opus.
These achievements reveal that the model is suitable not only for daily conversations or basic text production, but also for professional and academic use scenarios. The O3-pro is also a powerful tool for research-based work thanks to its capacity to work with comprehensive data sets and the ability to analyze step by step.