According to the AI company, it trained the “OpenAI o1 model” to spend more time thinking about problems before they respond, like a person would. Through training, they learn to refine their thinking process, try different strategies, and recognize their mistakes.
The new AI model can be used by healthcare researchers to annotate cell sequencing data, by physicists to generate complicated mathematical formulas needed for quantum optics, and by developers across fields to build and run multi-step workflows.
“We have developed a new series of AI models designed to spend more time thinking before responding. “They can reason through complex tasks and solve more difficult problems than previous models in science, coding and mathematics,” the company added.
In tests, the model performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology.
“We also discovered that he excels in math and coding. In a qualifying exam for the International Mathematics Olympiad (IMO), GPT-4o correctly solved only 13 percent of the problems, while the reasoning model scored 83 percent,” OpenAI said.
Coding skills were tested in contests and reached the 89th percentile in Codeforces competitions.
As a starter model, it still doesn't have many of the features that make ChatG useful, such as browsing the web for information and uploading files and images.
However, for complex reasoning tasks, this is a significant advance and represents a new level of AI capability.
"With this in mind, we will reset the counter to 1 and call this series OpenAI o1," the company said.
It has also developed a cheaper model of the "reasoning" series, called OpenAI o1-mini, which is a faster reasoning model and particularly effective in coding.
As a smaller model, o1-mini is 80 percent cheaper than o1-preview, making it a powerful and cost-effective model for applications that require reasoning but not extensive knowledge of the world.
The new AI model can be used by healthcare researchers to annotate cell sequencing data, by physicists to generate complicated mathematical formulas needed for quantum optics, and by developers across fields to build and run multi-step workflows.
“We have developed a new series of AI models designed to spend more time thinking before responding. “They can reason through complex tasks and solve more difficult problems than previous models in science, coding and mathematics,” the company added.
In tests, the model performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology.
“We also discovered that he excels in math and coding. In a qualifying exam for the International Mathematics Olympiad (IMO), GPT-4o correctly solved only 13 percent of the problems, while the reasoning model scored 83 percent,” OpenAI said.
Coding skills were tested in contests and reached the 89th percentile in Codeforces competitions.
As a starter model, it still doesn't have many of the features that make ChatG useful, such as browsing the web for information and uploading files and images.
However, for complex reasoning tasks, this is a significant advance and represents a new level of AI capability.
"With this in mind, we will reset the counter to 1 and call this series OpenAI o1," the company said.
It has also developed a cheaper model of the "reasoning" series, called OpenAI o1-mini, which is a faster reasoning model and particularly effective in coding.
As a smaller model, o1-mini is 80 percent cheaper than o1-preview, making it a powerful and cost-effective model for applications that require reasoning but not extensive knowledge of the world.
