By Emir Yildirim
ISTANBUL (AA) – The US-based ChatGPT-maker OpenAI announced on Thursday that it is introducing a series of new models named “o1-preview” and “o1-mini,” capable of spending “more time thinking before they respond.”
OpenAI said that the new series go through complex tasks in many areas, ranging from science to math much better than previous iterations of the firm’s chatbot, giving responses akin to PhD students from their respective domains, such as in physics, chemistry, and biology.
The firm stated the new models excel at math and computer programming in particular, as the new “reasoning model” scored 83% compared to the GPT-4o’s measly 13% at the International Mathematics Olympiad (IMO) qualifier.
Although lacking in features ChatGPT offers, such as scouring the worldwide web to fetch information to generate more up to date responses, complex tasks requiring deeper reasoning seen in the o1 models opens up a new era in artificial intelligence (AI) capability, according to OpenAI.
The o1-mini is focused on for its efficiency, as the firm hails it faster and cheaper by 80% compared to the o1-preview– reasoning model “particularly effective at coding.”
While access to the o1-preview will be exclusive to ChatGPT Plus members, the o1-mini will be included in the free tier membership, which requires a simple sign-in, while the currently most up-to-date iteration of ChatGPT, the 4o, will remain in service as is.