OpenAI released o3-mini, a reasoning model strong in science, maths and coding that it said responded 24% faster than o1-mini. It replaced o1-mini in ChatGPT and, for the first time, put a reasoning model in the hands of free users, subject to rate limits; Plus, Team and Pro users got higher limits, with Pro unlimited. It landed while DeepSeek R1, released on 20 January, was drawing attention, and was widely read as a response to it.