On April 16, 2025, OpenAI released o3 and o4-mini as their most advanced reasoning models. The o3-mini had been released earlier on January 31, 2025. OpenAI later followed up with o3-pro on June 10, 2025, described as their most capable model to date.
These models represent a significant step change in ChatGPT's capabilities for tasks requiring deep reasoning. They significantly outperform previous models in mathematics, coding, and scientific research.