OpenAI has officially launched o3-pro, a new artificial intelligence model the company is hailing as its most advanced to date. The latest addition to the o3 family is now available to ChatGPT Pro and Team subscribers, with rollouts for Enterprise and educational users expected in the coming days.
The o3-pro model, designed to handle complex reasoning tasks with greater precision, is intended to enhance performance across a range of professional fields, including science, education, mathematics, programming, and writing. OpenAI says the model surpasses its predecessor, o1-pro, in multiple domains—both in internal evaluations and user preference tests.
In a statement, OpenAI confirmed that o3-pro will also be accessible via its API. The company has set the pricing at $20 per million input tokens and $80 per million output tokens. The announcement follows a recent 80 per cent cost reduction in o3 token usage, a move aimed at broadening access to the platform’s latest capabilities.
According to internal testing by OpenAI, human reviewers consistently favoured o3-pro over previous models. It was preferred in 66.7 per cent of personal writing evaluations and in 62.7 per cent of coding tasks. Reviewers also rated it higher in areas such as clarity, adherence to instructions, and the depth of its responses.
The model is equipped with advanced tools that allow it to perform more sophisticated tasks, including web browsing, data file analysis, Python code execution, and visual reasoning. It also makes use of user memory for more tailored interactions. However, these functionalities can result in slightly longer response times compared to earlier models—a trade-off OpenAI suggests is worthwhile for users prioritising accuracy over speed.
Despite the advancements, the company acknowledges a few current limitations. The model does not support image generation, and temporary chats within ChatGPT are currently disabled due to a technical issue. Moreover, the Canvas workspace feature remains unavailable when using o3-pro.
OpenAI says o3-pro has demonstrated strong results in internal benchmark assessments, though comprehensive comparisons with rival AI models have not yet been published.
