The o3-mini is now set to surpass o1 in coding and reasoning tasks while reducing costs and latency, according to OpenAI.