OpenAI has released a new series of models, OpenAI o1
OpenAI o1 solves Olympiad-level math problems (AIME) at the level of the top 500 students in the USA and exceeds PhD-level accuracy in physics, biology, and chemistry tests (GPQA).
OpenAI o1 will be available today in ChatGPT and for select API users.
Information from the announcement:
One of the key features of o1 is its ability to model human thinking. The models were trained to spend more time analyzing problems, allowing them to find more accurate and creative solutions. This is especially noticeable in areas such as science, programming, and mathematics. For example, o1 successfully handles tasks from international math Olympiads, which were previously considered inaccessible to artificial intelligence.
Who is the o1 model intended for?
Scientists: The model can be used to analyze large volumes of data, create complex models, and generate new hypotheses.
Programmers: o1 can assist in writing code, debugging programs, and creating new algorithms.