According to the official blog of OpenAI, OpenAI has launched an artificial intelligence model o1 (internal code name "Strawberry") with reasoning ability. The model excels in complex task reasoning and solving scientific, coding, and mathematical problems. Tests show that o1 is at the same level as a doctoral student in challenging benchmark tasks such as physics, chemistry, and biology. In the International Mathematical Olympiad (IMO) qualifying exam, o1's correct answer rate reached 83%, far exceeding GPT-4's 13%. In the Codeforces programming competition, o1's performance exceeded 89% of the contestants.
OpenAI has also launched a faster and more economical version of o1-mini, which focuses on coding tasks and is 80% cheaper than o1-preview. Although the o1 series lacks some of the features of ChatGPT, it represents a new level of AI capabilities in complex reasoning tasks and can be widely used in fields such as medical research, physics, and software development. OpenAI plans to open o1-mini access to all free ChatGPT users.