The o3-pro model became available to ChatGPT Pro and Team users Tuesday, replacing the o1-pro model12. Enterprise and Education users will gain access next week, while developers can access o3-pro through OpenAI's API at $20 per million input tokens and $80 per million output tokens1.

OpenAI CEO Sam Altman expressed surprise at the model's performance, posting on X: "it is really smart! i didnt believe the win rates relative to o3 the first time i saw them"2. According to OpenAI, reviewers preferred o3-pro over both o3 and o1-pro across all tested categories, including science, programming, education, business, and writing12.

The model's primary advantage lies in enhanced reliability, particularly on OpenAI's "4/4 reliability" test, which requires correct answers to the same question across four separate trials2. This consistency comes with longer response times, as o3-pro accesses more tools including web search, Python, file analysis, and memory personalization2.

OpenAI simultaneously slashed o3 pricing from $10 to $2 per million input tokens and from $40 to $8 per million output tokens12. The move follows Google's Gemini 2.5 Pro pricing, which ranges from $1.25 to $2.50 for input tokens, according to The Decoder1.

"Until now, Gemini had been much cheaper than o3 at similar performance," The Decoder reported1. The o3-pro model is 87% cheaper than its predecessor o1-pro, according to Maginative3.

Several features remain unavailable at launch. Temporary chats have been disabled due to unresolved issues, and unlike other OpenAI models, o3-pro does not support image generation or the Canvas platform1. The company has not announced when these creative tools might be added.

OpenAI has positioned o3-pro for users who prioritize accuracy over speed, targeting professionals requiring detailed, high-stakes responses1.