OpenAI o1 Reasoning Model

OpenAI's o1 model introduced chain-of-thought reasoning as a training objective, producing models that could spend seconds to minutes thinking before answering complex problems. o1 scored in the 89th percentile on competitive programming problems, achieved PhD-level accuracy on physics benchmarks, and ranked in the top 500 of USA Math Olympiad qualifiers — surpassing virtually all human competitors.