Openai Launches O3-pro, Its Most Advanced Reasoning Model Yet

OpenAI has officially released o3-pro, an precocious AI reasoning exemplary that nan institution claims is its astir tin to date. The new o3-pro is being rolled retired to ChatGPT Pro and Team users, replacing nan earlier o1-pro model. Access for Enterprise and acquisition users is expected to travel wrong a week.

The caller exemplary is portion of nan o3 family, which OpenAI introduced earlier this year, and is designed to execute much reliably successful analyzable reasoning tasks, peculiarly successful fields specified arsenic science, education, mathematics, programming, and writing. The motorboat follows an 80% value simplification for o3 input and output costs, arsenic announced by CEO Sam Altman.

OpenAI says o3-pro is now accessible via some ChatGPT and its API. The exemplary is priced astatine $20 per cardinal input tokens and $80 per cardinal output tokens via nan API.

In soul evaluations, OpenAI reports that o3-pro outperformed its non-pro counterpart successful respective captious areas. Human testers preferred o3-pro successful 66.7% of individual penning tasks and 62.7% of machine programming tasks, according to nan company. Reviewers besides rated it higher for clarity, instruction-following, and comprehensiveness.

The exemplary is besides tin of utilizing devices that heighten its functionality. These see web search, record analysis, Python codification execution, machine imagination pinch reasoning, and entree to personification representation for much personalised responses. However, OpenAI warns that owed to nan usage of these tools, o3-pro responses whitethorn return somewhat longer compared to o1-pro. The institution advises users to prioritise nan exemplary erstwhile accuracy is much captious than speed.

Despite nan improvements, o3-pro comes pinch definite limitations. The exemplary cannot make images, and impermanent chats wrong ChatGPT are presently abnormal owed to a method issue. Additionally, nan Canvas workspace characteristic is not supported by nan model.

According to OpenAI, o3-pro has achieved beardown capacity successful soul benchmark testing. However, OpenAI has yet to merchandise broad head-to-head benchmark information comparing o3-pro pinch its apical competitors.

The motorboat of o3-pro further positions OpenAI successful nan competitory scenery of high-performance AI models, arsenic nan institution continues to refine its offerings for some wide and master usage cases.

Published on: Jun 11, 2025 6:23 PM IST