OpenAI, the company behind ChatGPT, introduced a new artificial intelligence model on Thursday, claiming it can "reason through complex tasks" and tackle more difficult problems than previous models.
The new model, named OpenAI o1, is part of a fresh series and has been shown to perform at a level similar to Ph.D. students in rigorous tests for physics, chemistry, and biology, according to a company blog post.
“We've trained these models to take more time analyzing problems before providing answers, much like a human would,” OpenAI explained. “Through this training, they’ve learned to refine their thought processes, try different approaches, and identify mistakes.”
Mira Murati, OpenAI's chief technology officer, described the launch as marking “the beginning of a new era in AI.”
“We are no longer confined to the pretraining paradigm; with the ability to scale through inference compute, we’re unlocking new capabilities and alignment potential,” Murati shared in a post on X, the platform formerly known as Twitter.
While the newly released o1-preview model offers enhanced reasoning for complex tasks, OpenAI acknowledged that GPT-4o, its last release, might still be better suited for everyday applications.
The o1-preview lacks some features that made earlier models more versatile, such as web browsing, file uploads, and image processing, according to the company.
“o1 is still flawed and limited, and while it may seem impressive at first glance, its shortcomings become more apparent with extended use,” OpenAI CEO Sam Altman noted in a post on X.
Despite the limitations, Altman expressed his pride in the company's achievements, calling the development of o1 a "monumental effort across the entire organization."
In addition, OpenAI introduced o1-mini, a faster and more affordable model designed for coding tasks, which is 80 percent cheaper than the o1-preview.
Both models, o1-preview and o1-mini, were made available to ChatGPT Plus and Team users on Thursday. ChatGPT Enterprise and Edu users are expected to gain access to the new models next week.