Openai It makes a smaller and more efficient version of its smartest artificial intelligence The available model is free as it seeks to answer the noise and enthusiasm revolves around a new open source offer from the Chinese company Deepseek AI.
I mentioned this before Openai was preparing for the new modelIt is called O3-MINI, for release on January 31. The company’s researchers are working on the additional work to get its willingness to peak time, according to the sources that spoke on the condition of anonymity.
O3-MINI, which bothered Openai in December, is a smaller version of the model that features the most advanced thinking capabilities in any Openai’s offer so far. The model can shatter difficult problems into parts formed in order to know the best way to solve them.
The company said: “This strong and rapid model develops the limits of the small models that it can achieve.” In a blog post Announcing the availability of O3-MINI.
Openai allows O3-MINI available to all Team Plus, Team and Pro of Chatgpt. The company says that the free version of Chatgpt will also be able to experience the O3-MINI but they will not be able to send the largest possible number of inquiries.
Openai is clearly using doctoral students to help train a new model for some time. Several weeks ago, the company began to employ PhD students at a price of $ 100 per hour to obtain “research cooperation” “that involves working on unpublished models,” according to an email that WIRED saw.
Openai seems to have employed doctoral students with experience in other areas through a company called MERCOR which you use regularly to find employees for typical training. accident Publishing job From Mercor on LinkedIn states: “The general goal of this project, which may become part of it is to create difficult scientific coding questions designed to test the capabilities of large language models in generating programming instructions to solve realistic scientific research problems.”
Publishing continues to give a problem with an amazing example similar to a problem in a standard called SCICODE designed to test the ability of the big language model to solve complex scientific problems.
The news comes Deepseek’s R1 Rail continues the American technology industry. The fact that this powerful model can be released free of charge puts the pressure on Google and Anthropor to reduce their prices.
Openai is especially eager to prove that it is still at the forefront of developing and marketing artificial intelligence, according to the sources within the company.
The DEPSEK model is freely available in innovations that made it more efficient for both train and service. The company seems to have developed it with much lower resources than Openai and other American companies that are currently building the border AI models, although the fine details of Deepseek are still unknown. Openai He says he believes R1 may have merged out of its models in its training.
Do you have advice?
Are you a current or former employee in Openai? We would like to hear from you. Using a phone or computer that is not action, the connection will be made in will_knight@wired.com Or on the Wak01 username.
The latest Openai R1 may not exceed the price, but it shows that the company will make a part of its concentration to move forward. Openai also says that the model is particularly strong in mathematics, science and coding.
The company says that the latest model will also include new features, including the ability to take advantage of the web searches, communication functions from the user code, and switch between different levels of thinking that exceed the speed of problem solving capabilities.
Dibsic’s sudden rise also raised questions about the US government’s strategy to reduce China’s rise in artificial intelligence. The past two US administrations have introduced a number of sanctions to reduce China’s ability to reach the most advanced NVIDIA chips used to build advanced AI models. Deepseek described several types of NVIDIA chips in their research, but it is still not clear exactly what has been used.
