Open O3-MINI unveiled the Open

The OpenIA unveiled the new O3-Mini model last night, which the company said, significantly reduced the cost of reasoning than the O1 model, while also providing more speed and accuracy.

“O3-Mini is fast and intelligent and is available to users through API,” said Sam Altman, CEO of OpenA. This model is capable of searching the Internet and displaying its thinking process, a feature that can have a great impact on complex problems.

Focus on developing more advanced models

After the successful release of O3-Mini, which many consider to be superior to O1, Open has focused on developing the larger version of the model, O3.

The model also performs brilliantly in the “Last Humanity Exam” test, which is one of the most difficult challenges of artificial intelligence. O3-Mini (Medium) tops the table at a 5 % accuracy, while O3-Mini (HIGH) ranks second in the second place and the R1 model of Deepsic with a accuracy of 4.9 % accuracy. .

Researchers believe that if artificial intelligence can reach an acceptable score in this test, it can be seen as a potential threat to human superiority in some areas.

Increase accuracy in calculations and reasoning

OpenAI has announced that the Medium version of the model has similar O1 in areas such as mathematics, coding and functional science, while O3-Mini has been optimized for STEM argument. According to experts, this model has better responses and ability to argue better than the O1-Mini version.

According to the company, O3-Mini has a 5 percent less mistake in the difficult real world issues than its previous model, while also providing responses at a speed of 5 %.

Users and developers’ access to O3-Mini

Since last night, the O3-Mini model has been available to ChatGpt users. Free users can use this model through the new “Reason” button in chat, while ChatGpt Plus and Team users have a daily use limit. ChatGpt Pro subscribers will have unlimited access to this model.

The model will also be available to ChatGpt Enterprise and ChatGpt Edu users for about a week.

The API Open has also released this model for developers, though it does not currently support image analysis. Developers can adjust the “effort to reason” to adjust the model to the bottom, medium or high to manage its thinking level.

Reduce processing costs

The O3-MINI model costs $ 1.5 for one million input token and $ 1.2 for one million output token, which is about 2 percent cheaper than the O1-MINI. The price is in the direct competition with the Deepsic R1 model, which set the cost of one million input and outlet token at $ 1.2 and $ 1.2, respectively.

By default, ChatGPT has set this model in moderate reasoning, which, according to OpenII, offers a “balance of speed and accuracy”. However, money users can use the O3-Mini-High version for deeper thinking, although it is lower processing speed, but provides more accurate outputs.

In all versions, the model supports internet search to access up -to -date data, but Open -AI has emphasized that it is still a prototype and may need to be improved in some cases.