TechFlow news, according to OpenAI's official blog, OpenAI has launched the o1 AI model (internal codename "Strawberry"), an artificial intelligence model with reasoning capabilities. The model excels in complex task reasoning and solving challenges in science, coding, and mathematics. Testing shows that o1 performs on par with PhD students in challenging benchmarks across physics, chemistry, and biology. In the International Mathematical Olympiad (IMO) qualifying exam, o1 achieved an 83% correct solution rate, far surpassing GPT-4's 13%. In Codeforces programming competitions, o1 outperformed 89% of participants.
OpenAI has also introduced a faster and more cost-effective version, o1-mini, focused on coding tasks and priced 80% lower than o1-preview. While the o1 series currently lacks some features of ChatGPT, it represents a new height in AI capability for complex reasoning tasks and can be widely applied in fields such as medical research, physics, and software development. OpenAI plans to grant all free ChatGPT users access to o1-mini.
