Chinese company DeepSeek has announced the release of an improved version of its artificial intelligence model R1, featuring 685 billion parameters and available under the open MIT license.
G. Ostrov
Chinese technology startup DeepSeek has announced the release of a modernized version of its AI model R1, now available on the Hugging Face platform under the free MIT license. According to the company's official statement on the WeChat social network, the new version has received several improvements and can be freely used for commercial purposes.
Currently, the Hugging Face repository lacks detailed technical documentation for the model. Only configuration files and weight parameters are available — numerical values that determine the functionality and capabilities of the AI system. The updated R1 model contains 685 billion parameters, requiring significant computational resources. Experts note that without special optimization, running such a model on standard personal computers is practically impossible.
It's worth recalling that the DeepSeek project attracted considerable attention from the global community at the beginning of this year after presenting the first version of R1, which managed to provide serious competition to OpenAI's developments. However, the startup's achievements raised concerns among American regulatory bodies, who view the company's technologies as a potential threat to US national security.
Despite geopolitical challenges, DeepSeek continues the active development of its AI platform. The use of the open MIT license provides developers and commercial organizations with the opportunity to freely test and integrate the R1 model into their solutions, although powerful computing systems are required for its full operation.
For more detailed information about the R1 model, we recommend visiting the official DeepSeek website.
In case of any problems, contact us, we will help quickly and efficiently!