OpenAI has introduced a new generative AI model called “o3-pro“, which the company claims is the most powerful in their lineup. The “o3-pro” model has replaced the previous “o1-pro” for ChatGPT Pro and Team users, and it will be available to corporate and educational clients next week. Developers can also access “o3-pro” through the API, with a cost of twenty dollars per million input tokens and eighty dollars per million output tokens.
According to internal testing, experts have highlighted “o3-pro” in areas such as science, education, programming, business, and writing assistance. The model received high marks for clarity, completeness, accuracy, and its ability to follow instructions clearly. It can work with various tools — searching the internet, analyzing files, processing images, using Python, and personalizing responses thanks to memory.
Among the limitations — “o3-pro” currently does not create images, does not support the Canvas feature, and temporary chats in ChatGPT are unavailable while the company resolves technical issues. The model’s responses are generated slightly slower than the previous version, but this is compensated by the quality of the results.
OpenAI’s internal tests show that “o3-pro” outperforms competitors on popular math and science tests. Specifically, on AIME 2024, the model performed better than Google’s “Gemini 2.5 Pro,” and on GPQA Diamond, which assesses PhD-level science knowledge, it surpassed Anthropic’s “Claude 4 Opus.”
