OpenAI introduced a new AI model on Thursday, named o1, which the company claims can "reason through complex tasks and tackle tougher problems" compared to its predecessors. The o1 model, part of the new OpenAI o1 series, is reported to perform at a level comparable to Ph.D. students on challenging benchmark tests in physics, chemistry, and biology. According to OpenAI's blog, this model has been designed to think through problems more thoroughly before responding, much like a human would. It learns to refine its problem-solving strategies and correct its mistakes through training.

// Cras eget sem nec dui volutpat ultrices.

Mira Murati, OpenAI's Chief Technology Officer, described this model as the beginning of a "new era in AI," emphasizing that it transcends previous limitations by leveraging advanced inference compute to unlock new capabilities and alignment.

Although the new o1-preview model exhibits enhanced complex reasoning skills, OpenAI noted that the previous GPT-4o model might still be more effective for everyday tasks. The o1-preview model lacks some features found in earlier models, such as web browsing and file/image uploads.

OpenAI CEO Sam Altman acknowledged the model's current limitations but expressed pride in the team's achievements. In addition, OpenAI released o1-mini, a faster, more cost-effective model specialized in coding, which is 80% cheaper than o1-preview.

// Cras eget sem nec dui volutpat ultrices.

The o1-preview and o1-mini models were made available to ChatGPT Plus and Team users on Thursday, with ChatGPT Enterprise and Edu users expected to gain access next week.