Openai is creating a smaller, more effective version of it Artificial intelligence The model is available for free because it tries to answer the exaggeration and enthusiastic around a new open source from China's Startup Deepseek.
Wired previously reported that Openai is preparing a new modelcalled O3-mini, to release on January 31.
O3-mini, Openai teased in December, which is a smaller version of the model with the most advanced AI reasoning of any Openai proposal so far. The model can divide difficult problems into constituent parts to find the best way to solve them.
This powerful and fast model promotes the boundaries of what the small models can achieve, the company said In a blog post Notice of O3-Mini's comfort.
Openai is providing O3-mini for all professional communities, groups and users of TATGPT. Users of the free version of chatgpt will also be able to try O3-mini but will not be able to send more queries, the company said.
Openai clearly used PhD students to help train a new model for a while. A few weeks ago, the company began recruiting PhD scientific students at $ 100 per hour for a research cooperation of Muslims, which involved working on models. Not yet released, according to an email viewed by Wired.
Openai also seemed to have recruited a doctorate with expertise in other areas through a company called Mercor that they often use to find employees to train the model. A recent Post From Mercor on LinkedIn: The overall goal of this project that you can become a part of creating challenging scientific encryption questions is designed to check the ability of the language models. Large language in creating codes to solve practical scientific research issues.
The posting of work continues to give an example problem similar to a problem in a benchmark called Scicode designed to check the ability of a large language model to solve complex scientific problems. .
News came Deepseek's R1 Continue to shake the US technology industry. The fact that such a powerful model can be released free of charge on Google and anthropological to reduce their prices.
Openai is particularly eager to prove that it is always at the forefront of the development and commercialization of AI, according to the sources inside the company.
Deepseek's free available model combines innovation to make it more effective for both training and service. The company seems to have developed by using less resources than Openai and other US companies currently building frontier AI models, although the exact details of DeepSeK's spending have not yet been know. Openai Say it believes R1 may have combined its output from its models into its training.
There is a tip?
Are you an existing employee or former employee at Openai? We want to hear from you. Using phones or computers does not work, contact Knight at Will_knight@wired.com Or on the signal through its user name Wak01.
The latest model of Openai may not overcome the R1 in terms of price, but it shows that the company will make the effectiveness of its focus in the future. Openai also said that a particularly powerful model in math, science and encryption.
The company says that the latest model will also combine new features, including the ability to exploit searches on the web, call the functions from the user code and convert between different theory levels to fight Change speed for problem solving capabilities.
Deepseek's sudden increase has also asked questions about the US government strategy to curb China's rise in AI. The two previous US governments issued a number of sanctions to curb China's capabilities to access the most advanced NVIDIA chips that are often used to build advanced AI models. Deepseek has described some of the NVIDIA chips in his research, but it is still unclear what has been used.