OpenAI previews o1, a model trained to spend more time reasoning
OpenAI releases o1-preview, shifting frontier-model competition toward reinforcement learning and additional computation at inference time.
What happened
Event details
OpenAI o1-preview was the first public model in a new reasoning series trained to work through difficult problems before answering. OpenAI emphasized performance in mathematics, coding, and science, while the initial preview had higher latency, narrower product features, and limited availability compared with general ChatGPT models. The release popularized test-time compute as a new scaling axis: better answers could come not only from a larger pretrained model, but from spending more computation on each problem.
Assessment
Why it matters
o1 made reasoning-time computation a mainstream product category and set the competitive context for open models such as DeepSeek-R1.