Jakarta, INTI – On Tuesday, June 10, 2025 US time, OpenAI officially introduced o3-pro, their latest AI model which is claimed to be the most advanced reasoning model the company has ever released. The model is currently available for ChatGPT Pro and ChatGPT Team users, and has been released via API for developers. o3-pro is not just another upgrade, it brings major improvements in the accuracy, completeness, and clarity of AI responses, especially in complex tasks such as programming, science, math, and writing.
Significant Improvement from the Previous Generation
The o3-pro model is an enhancement of the “o3” reasoning model previously introduced in April 2025. Compared to its predecessor, o3-pro comes with a number of advantages: it is more reliable, follows instructions better, and provides clearer and more complete answers.
One method of testing its reliability is called “4/4 reliability”, which is the AI's ability to correctly answer the same question four times in a row. This shows that o3-pro is not only intelligent, but also consistent.
Deeper Reasoning Skills and More Precise Responses
This model is perfect for use in tasks that demand high precision, such as summarizing, analyzing data, answering questions that require deep understanding, or composing long, structured explanations.
In internal benchmark tests, o3-pro managed to outperform many leading AI models. In the AIME 2024 math test, o3-pro surpassed Google's Gemini 2.5 Pro. In PhD-level science, o3-pro also surpassed Anthropic's Claude 4 Opus model.
Not just a quick answer, but the right answer
According to OpenAI, the o3-pro model excels in delivering.structured and easy-to-understand answers, including when asked to adapt certain language styles, formats, or constraints. However, due to the complexity of the reasoning process and the tools used, o3-pro's response time is slower than models like o1-pro. Therefore, OpenAI recommends that users use o3-pro for important questions that require precision, rather than for quick answers.
Advanced Features, But Still Limitations
The o3-pro supports tools such as visual understanding, file analysis, Python programming, as well as the use of user memory for more personalized responses. However, this model does not yet support image creation, unlike the GPT-4o or o4-mini. Also, the Canvas feature and temporary chats are not yet available.
Already Available in API, Ready to Test in Various Sectors
For developers, o3-pro has been available on the OpenAI API since June 10, 2025. It is priced at USD 20 per million input tokens and USD 80 per million output tokens, equivalent to around Rp 325,000 and Rp 1.3 million. This model is also scheduled to be available to ChatGPT Enterprise and ChatGPT Edu users the following week.
For information, OpenAI deliberately skipped naming “o2” to avoid potential brand conflicts with British telecommunications company O2. Previously, the o1 model or known as Strawberry was launched last September.
Conclusion
The launch of o3-pro marks a major leap in the development of high-reasoning AI models. With improved reliability, sharper instruction comprehension, and clearer and more complete responses, it is an ideal solution for a variety of complex needs in research, education, and data analysis. Although it still has some limitations, o3-pro shows a new direction in the evolution of artificial intelligence that is not only fast, but also precise and reliable.
Read More:Pandemic Predictions Now More Accurate Thanks to AI by US Scientists