OpenAI is slowly inviting selected users to test a whole new set of reasoning models named o3 and o3 mini, successors to the o1 and o1-mini models that just entered full release earlier this month.

Simply put, reasoning models are trained to "think before they speak," meaning they take more time to process the prompt but provide higher-quality responses. As a result, like older models, o3 and o4 ...