OpenAI just launched its latest model, but this time, it has not been branded with a "ChatGPT" or "GPT" name. Instead, it's naming the model after the company itself: OpenAI o1.
In a blog post, OpenAI described how these new models work. They sound like the rumored features that were previously reported under the code name "Strawberry." The OpenAI o1 models have been created to, in the company's words, "spend more time thinking before they respond."
The blog post added that the goal is to have OpenAI o1 work on much more complex tasks and projects:
In our tests, the next model update performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology. We also found that it excels in math and coding. In a qualifying exam for the International Mathematics Olympiad (IMO), GPT-4o correctly solved only 13% of problems, while the reasoning model scored 83%. Their coding abilities were evaluated in contests and reached the 89th percentile in Codeforces competitions.
OpenAI o1 cannot do some of the things that its ChatGPT LLMs can do, such as finding information on the web or checking out uploaded images and other files. Nevertheless, the company feels that this new model is a major step forward in generative AI, which is why it named the model OpenAI o1, which stands for "resetting the counter back to 1."
In addition to the main OpenAI o1 model, the company is also launching OpenAI o1-mini. It has been designed to be faster and cheaper to use and concentrates on helping with coding tasks.
Subscribers to both ChatGPT Plus and ChatGPT Team will be able to access the new OpenAI o1 models starting today. The models will be available next week for subscribers to ChatGPT Enterprise and ChatGPT Edu plans. In the near future, the OpenAI o1-mini model will be accessible to free ChatGPT users.
0 Comments - Add comment