o1-preview — reasoning models arrive

Chain-of-thought reasoning at inference time. New paradigm: 'thinking' models that trade speed for accuracy.

More context