OpenAI o1 and o1 Mini models introduced, these AIs think before answering

OpenAI announced a new AI series on Thursday. The company has introduced the OpenAI o1 series which is a large language model designed to solve difficult problems. The company says that it comes with better reasoning capability due to continuous learning.

The preview of this model has just been released. The company says that this AI model thinks before answering and because of this it is able to solve difficult reasoning problems. This AI model is more effective than previous models in the areas of science, math and coding.

How does OpenAI o1 work?

The company says that OpenAI o1 is able to solve even difficult problems by using reinforcement learning. This AI model thinks a little before giving any answer and corrects its mistakes. In the test, the performance of OpenAI o1 matched that of PhD students of Physics, Chemistry and Biology. Not only this, its performance also matched that of students of Mathematics and Programming.

Also read: OpenAI ChatGPT was down worldwide, now the service is back on track, what did the company say?

This AI model beats GPT-4o in almost all reasoning heavy tasks. With just one sample, it scored 74 percent in the AIME math exam. Whereas after 64 samples its score reached 83 percent and after using re-ranking the score reached 93 percent.

It has also surpassed human PhD level performance in the GPQA Diamond benchmarks of Physics, Chemistry and Biology. Its score in the International Olympiad of Informatics 2024 was 213 points and rank was 49th. This AI performs better with strategic test time submission.

Also read: OpenAI brings search engine, know the specialty of SearchGPT

The company has said that OpenAI o1 thinks a lot before solving a difficult problem, so it can solve that problem easily by dividing it into smaller steps. It is constantly improving with more training time.

Which users will get access to this?

OpenAI has designed this o1 model to solve complex problems. Its focus is in the areas of science, coding and mathematics. It can be used to generate mathematical formulas for quantum optics and execute multi-step workflows in development.

Only premium users will get access to this. ChatGPT Plus and Team users will get access to both OpenAI o1 and OpenAI o1 Mini. Whereas for ChatGPT Enterprise and Education users, this will be available from next week.