o1 Preview

Research preview of the o1 series of models, trained with reinforcement learning to perform complex reasoning. o1 models think before they answer, producing a long internal chain of thought before responding to the user.