After much speculation over the past few weeks, the much-speculated Project Strawberry has finally been announced by OpenAI, the company that developed ChatGPT. These AI models, called o1 and o1-mini, are considered advanced prototypes that can solve complex problems with human-like logic, albeit slower.

The most complete version, OpenAI-o1, differs from other existing models, such as the powerful GPT-4o, in that this AI is designed to think in a similar way to humans, rather than just writing canned answers. The company claims that through this process, it can improve the logic of the models, create different response strategies, and understand their mistakes.

According to Mira Murati, OpenAI’s chief technology officer, new projects are developed using effortful learning, which means models receive positive or negative feedback on whether their answers are correct, improving their reasoning abilities.

The new OpenAI-o1 and o1-mini are complementary technologies, not successors to GPT-4o.

o1 is ideal for natural sciences and technology

In the company’s tests, OpenAI-o1 codes perform better in cases related to mathematics, biology, physics and chemistry. In an advanced math test called AIME (American Invitational Mathematics Examination), the model got 83% of the questions correct, while GPT-4o got only 13% correct.

On the other hand, o1-mini is an AI model with reduced capabilities but focused on energy efficiency and code generation. Despite the advancements, OpenAI points out that the versions have flaws, as they cannot search the internet and also cannot analyze images or files.

Demonstration with complex mathematical resolutions.
Demonstration with complex mathematical resolutions.

ChatGPT Plus or Team users will be able to use the o1 model starting today and select the new project from the ChatGPT selection tab. For now, o1 is limited to 30 messages per week, while o1-mini can respond to up to 50 messages. OpenAI plans to make o1-mini available to all free users of the platform in the future.

For developers, the path is much more expensive. The OpenAI-o1 model costs $15 for one million input tokens and $60 for one million output tokens.

Source: Tec Mundo

Previous articleLando Norris has shown that Formula 1 could have a McLaren made out of Lego
Next articleProjectors from R $ 115 on Shopee: 14 cheap options to have at home
I am a passionate and hardworking journalist with an eye for detail. I specialize in the field of news reporting, and have been writing for Gadget Onus, a renowned online news site, since 2019. As the author of their Hot News section, I’m proud to be at the forefront of today’s headlines and current affairs.

LEAVE A REPLY

Please enter your comment!
Please enter your name here