Newest OpenAI Reasoning Model Released
OpenAI is officially unveiling its latest model, the o3-mini, starting today, Friday, January 31st. The company shared this news in a blog on its website , just over a month after the model was officially announced during the 12 Days of OpenAI.
Like every updated generative AI model, the o3-mini is an improvement over the o1-mini, but not as much as you might think. OpenAI claims that both models perform equally well in math, programming, and science, but o3-mini offers faster responses to user queries—24% faster in A/B testing. According to the company, testers who compared the models found o3-mini gave “more accurate and clear answers, as well as stronger reasoning abilities.” And at “average reasoning effort,” o3-mini matches o1 in some assessments of reasoning and intelligence.
Like o1-mini, o3-mini is a reasoning model, a type of AI model that “thinks” about answers before reacting to them. o3-mini has three different reasoning “efforts” depending on the use case: low, medium and high. For example, on a math test, medium and high level reasoning o3-mini outperforms o1-mini, and high level even outperforms o1 (a more powerful version of o1-mini). All three projects outperform o1-mini on PhD-level science, but o1 beats them all.
o3-mini replaces the o1-mini model for all users. OpenAI doesn’t directly state why you can’t use the o1-mini in the future, but it does advertise that the o3-mini has higher speed limits and lower latency than the previous model.
At launch, only ChatGPT Plus, Team and Pro users can access o3-mini. OpenAI says enterprise users will be able to access the model in a week. (In addition, Plus and Team users will see their daily rate limit increase from 50 messages on o1-mini to 150 messages.)
However, free users will be able to try out o3-mini on a limited basis, either by selecting the “Reasoning” option in the message editor or by re-generating the answer. OpenAI says this is the first time free users have access to the reasoning model in ChatGPT, which comes a day after Microsoft offered o1 reasoning to Copilot users for free .
You can learn more about o3-mini in our post here . But since the model is just being rolled out today, we won’t know exactly how it works until real testers start using it.