Member-only story
Introducing OpenAI o1-preview: A New Era of Reasoning Models
On September 17, 2024, OpenAI introduced a new series of reasoning models called OpenAI o1-preview. These models are designed to solve difficult problems better than before, helping in areas like science, coding, and math.
What is OpenAI o1-preview?
The o1-preview models are special because they spend more time thinking before they give answers. This is similar to how a person thinks through problems. During training, these models learn to improve their thinking, try different methods, and see where they make mistakes.
In tests, these new models performed like PhD students on tough tasks in physics, chemistry, and biology. They are especially good in math and coding. For example, in a hard math exam for the International Mathematics Olympiad, the older model, GPT-4o, solved only 13% of problems, while the o1-preview scored an impressive 83%! Also, in coding competitions, o1 models reached the 89th percentile on Codeforces.
