Release anniversaries, 12 Sept 2027
OpenAI o1 preview
On 12 September 2024 OpenAI previewed o1, the first widely available reasoning model.
OpenAI released o1-preview and o1-mini on 12 September 2024. The models were trained with reinforcement learning to work through a hidden chain of thought before answering, and OpenAI showed that accuracy kept improving the longer they were allowed to think. They scored far higher than GPT-4o on competition maths, science questions and programming contests, at the cost of slower and more expensive answers. The full o1 followed in December 2024 and o3 in 2025. Rival labs soon released reasoning models of their own.