o3-pro (2025-06-10)

The o series models are trained using reinforcement learning and are capable of thinking before answering and carrying out complex reasoning. The o3-pro model uses more computing resources to think more deeply and consistently deliver high-quality responses. o3-pro is only available in the Responses API to support multi-turn model interactions before responding to API requests, as well as other advanced API features in the future. Because o3-pro is designed to solve tough problems, some requests may take several minutes to complete, and background mode can be used to avoid timeouts.