OpenAI makes a smaller, more efficient version of the smartest artificial intelligence model available for free because it wants to answer the hype and enthusiasm that is eagerly around a new open source range of the Chinese AI Startup Deepseek.
Wired previously reported that OpenAi prepared the new model, called O3-Mini, for release on January 31. The researchers of the company have worked overtime to prepare it for Prime Time, according to sources that spoke on the condition of anonymity.
O3-Mini, which was plagued in December, is a smaller version of the model with the most advanced AI reasoning options of every OpenAI offer to date. The model can break difficult problems in component parts to find out how they can best be solved.
“This powerful and fast model promotes the limits of what small models can achieve,” the company said in a blog post that announced the availability of O3-Mini.
OpenAi makes O3-Mini available for all Chatgpt Plus, Team and Pro users. Users of the free version of Chatgpt can also try O3-Mini, but cannot send so many questions, says the company.
OpenAi apparently uses PhD students to train a new model for a while. A few weeks ago, the company started recruiting the students of PhD student for $ 100 per hour for a “research cooperation” where “would work on non -depicted models”, according to an e -mail viewed by Wired.
OpenAi also seems to have recruited PhD students with expertise in other areas through a company called Mercor that regularly uses to find staff for model training. A recent vacancies from Mercor on LinkedIn states: “The general goal of this project of which you may be part of challenging scientific coding questions that have been designed to test the possibilities of large language models in generating the code for solving realistic scientific research problems. “
The vacancies also give by giving an exemplary problem that seems striking to a problem in a benchmark called Scicode that is designed to test the ability of a large language model to solve complex science problems.
The news is because Deepseek's R1 continues to roast the American technical industry. The fact that such a powerful model can be released for free, puts pressure on Google and anthropic to lower their prices.
OpenAi is particularly eager to show that according to sources within the company it is at the forefront of developing and commercializing AI.
The freely available Deepseek model contains innovations that made it more efficient for both training and serving. The company seems to have developed it with much fewer resources than OpenAI and other American companies that are currently building frontier AI models, although the precise details of the expenditure of Deepseek are unknown. OpenAi says it believes that R1 may have included the output of its models in its training.
Do you have a tip?
Are you a current or former employee at OpenAI? We would like to hear from you. With the help of a NonWork telephone or computer, contact Will Knight on will_knight@CBNewz or on Signal via his username WAK01.
The newest model from OpenAi may not exceed R1 in terms of price, but it shows that the company will make part of its focus in the future. 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 functions, including the possibility of using web search assignments, call functions of a user code and switching between different reasoning levels that trade in speed for problem -solving options.
The sudden rise in Deepseek has also asked questions about the strategy of the US government to curb the rise of China in AI. The last two American administrations have introduced a number of sanctions to curb China's ability to access the most advanced NVIDIA chips that are usually used to build advanced AI models. Deepseek described different types of Nvidia chips in his research, but it remains unclear what exactly was used.
