On Sept. 12, OpenAI revealed a demo of its new concept, OpenAI o1, designed to handle difficult things such as writing script, solving math problems and performing serious argument. It is the first of the long-rumored next-generation AI family known as” Strawberry”.
ChatGPT Plus, Team people, and engineers with OpenAI API use Tier 5 can now get the demo version of the full design, o1-preview.
These people can also get o1-mini — a smaller, faster version of the o1 design that is especially effective at programming. As a smaller model, the software giant says it is “80 % cheaper than o1-preview, making it a strong, cost-effective unit for applications that require logic but not broad earth information”.
Beginning next year, consumers of ChatGPT Enterprise and Edu may have access to both models, according to Start AI.
” We even intend to make o1-mini access available to all ChatGPT Free people,” the company stated in a press release.
below is o1, a series of our most worthy and aligned types however: https: //t. co/yzZGNN8HvD
O1 is still imperfect, limited, and still appears more impressive at first use than it does once you’ve used it for a while. pic. twitter.com/Qs1HoSDOz1
— Sam Altman ( @sama ) September 12, 2024
O1 requires more time to reason through more challenging problems.
Instead of furthering GPT-4’s language capability, OpenAI o1 and o1-mini focus on science, creating and debugging code and math. The model is shown in a demonstration video creating a playable game in the 1970s in the vein of Snake games. As OpenAI explained, o1 can be used by:
- Researchers in the field of health care to annotate cell sequencing data.
- Physicists work to create the complex mathematical equations required for quantum optics.
- Developers in all fields to build and execute multi-step workflows.
O1 placed in the 89th percentile on the competitive programming test Codeforces, according to OpenAI, and placed among the top 500 students in the country as a qualifier for the USA Math Olympiad.
By nature, o1 will take longer to answer than ChatGPT or GPT-4.
o1-preview can output a maximum of 32k tokens, while o1-mini can output a maximum of 64k tokens. Depending on the complexity of the text, a token can be as brief as one character or as long as one word.  , Both versions of the new model support text input only, not audio or images.
To help developers determine whether o1 is best for their work, OpenAI created a best practices guide.
In the model’s system card, where OpenAI outlines red-teaming efforts and other security considerations, o1 received a “medium” safety rating in two categories. The independent research group Apollo Research noted that O1 “has the fundamental skills needed to carry out simple in-context scheming,” which is the use of “gaming their oversight mechanisms as a means to achieve a goal.” On the other hand, the model gains a better understanding of safety regulations thanks to the model’s deeper reasoning.
