OpenAI launches o3-pro, its most advanced reasoning model yet

OpenAI has launched o3-pro, its latest advanced AI reasoning model, touted as the company's most capable yet. This new model is being introduced to ChatGPT Pro and Team users, effectively replacing the previous o1-pro version. Access for Enterprise and educational users is set to roll out within the next week, expanding its reach significantly. Part of the o3 family, which OpenAI debuted earlier this year, o3-pro is specifically engineered for improved reliability in intricate reasoning tasks across various domains, including science, education, mathematics, programming, and writing. This launch comes on the heels of a major 80% reduction in input and output costs for o3, as highlighted by CEO Sam Altman. Starting today, o3-pro is available for Pro and Team users in the model picker, marking a significant upgrade. Enterprise and educational users will follow shortly after. Utilizing the same foundational model as o3, details regarding safety measures can be referenced in the o3 system card. Both ChatGPT and its API now support o3-pro, with pricing set at $20 per million input tokens and $80 per million output tokens via the API. In internal assessments, o3-pro has shown superior performance compared to the non-pro version in critical areas. Notably, human evaluators favored o3-pro in 66.7% of personal writing tasks and 62.7% of programming tasks, with reviewers praising it for clarity, instruction adherence, and thoroughness. Moreover, o3-pro is equipped with tools that significantly enhance its capabilities, including web search, file analysis, Python code execution, and computer vision with reasoning. It can also recall user memory for more tailored responses. However, OpenAI cautions that the use of these tools may result in slightly longer response times compared to o1-pro, suggesting that users prioritize accuracy over speed when necessary. Despite its advancements, o3-pro does have limitations. It currently cannot generate images, and temporary chats within ChatGPT are disabled due to ongoing technical issues. Additionally, the Canvas workspace feature is not compatible with this model. OpenAI has reported impressive results from internal benchmark tests, but comprehensive comparisons with leading competitors have yet to be released. The introduction of o3-pro strategically positions OpenAI within the competitive realm of high-performance AI models as it continues to enhance its offerings for both everyday and specialized applications.