The model demonstrates PhD-level accuracy in solving scientific problems.
OpenAI has announced the release of a new series of AI models, codenamed Strawberry, capable of solving complex problems through logical reasoning. The key features of the new model are presented on the company’s official website.
The new model, named o1, is already available in ChatGPT and through the company’s API. A faster version, o1-mini, is also planned for release.
This innovation promises to significantly enhance the ability of artificial intelligence to analyze and solve tasks in the fields of science, programming, and mathematics.
The Strawberry series uses a technique called “chain of reasoning,” which enables the AI to break down complex problems into smaller logical steps. This greatly improves the AI’s efficiency in solving tasks in scientific, programming, and mathematical domains. The model is now also capable of deep reasoning and analysis.
The o1 model achieved an accuracy rate of 83% on the International Mathematical Olympiad qualifying exam. In comparison, the previous model, GPT-4o, only reached 13%.
“We trained these models to carefully analyze problems before giving an answer, just like humans do. In the process, they refine their strategies, learning to recognize and correct mistakes,” the OpenAI blog states.
