OpenAI has introduced o3-pro, a new AI model it describes as its most advanced to date. Built on the foundation of the o3 model, which debuted earlier this year, o3-pro represents a significant leap in performance. Unlike traditional AI systems, reasoning-based models like o3-pro tackle problems methodically, step by step, allowing for greater accuracy and consistency—particularly in complex areas such as mathematics, physics, and programming.
OpenAI announced that “o3-pro is now being made available to all Pro subscribers on ChatGPT as well as through the API.” The company explained that while o3-pro is based on the same core architecture as o3, it delivers improved results across key areas of performance.
“Like the base o3 model, o3-pro is equipped with powerful tools that make ChatGPT highly versatile—it can browse the internet, interpret images, work with files, run Python code, remember user preferences, and more. However, because it uses these advanced capabilities, its responses may take a bit longer to generate compared to o1-pro. For tasks where precision is more important than speed, this tradeoff is often worth it,” the company said in a statement.
The model has also met OpenAI’s stringent “4/4 reliability” standard, which requires it to provide accurate answers in all four trials of a given question—passing just once isn’t enough to qualify.
Much like the earlier o1-pro model, the latest o3-pro continues to demonstrate exceptional academic strength. “Similar to OpenAI o1-pro, o3-pro performs exceptionally well in subjects like mathematics, science, and programming, as evidenced by academic testing,” the company noted.
In the meantime, OpenAI has slashed the cost of its o3 model by 80%, bringing input pricing down to $2 per million tokens and output to $8 per million tokens. In contrast, the new o3-pro model is priced at $20 per million tokens for input and $80 per million tokens for output—making it 87% more affordable than the o1-pro model.
