$20/M input tokens$80/M output tokens$15.30/K input imgs$10/K web search
The o-series of models are trained with reinforcement learning to think before they answer and perform complex reasoning. The o3-pro model uses more compute to think harder and provide consistently better answers.