OpenAI releases o1, its first model with ‘reasoning’ abilities

OpenAI released o1, the first in a planned series of "reasoning" models trained to answer complex questions, alongside the smaller, cheaper o1-mini. According to the article, o1 does better than GPT-4o at writing code and solving multistep problems, though it is more expensive and slower to use. OpenAI told the article's author that o1 scored 83 percent on a qualifying exam for the International Mathematics Olympiad, compared with 13 percent for GPT-4o, and reached the 89th percentile in Codeforces programming competitions. ChatGPT Plus and Team users received access starting that day, while OpenAI's research lead, Jerry Tworek, said the model "hallucinates less" but that hallucinations are not solved.