Chinese startup DeepSeek has released an improved version of its R1 artificial intelligence model and made it available on the Hugging Face platform under an open MIT license. The company announced on WeChat that this updated model has undergone a minor enhancement and can be used freely in commercial applications.
The Hugging Face repository currently lacks a detailed description of the model. It only contains configuration files and “weights” – the numerical values that dictate how the model operates and what it can do. This updated R1 contains 685 billion parameters, making it require significant computing resources. As TechCrunch points out, it’s unlikely that typical personal computers could run such a model without substantial optimization.
Earlier this year, DeepSeek gained considerable attention with the initial release of R1, which was seen as a competitor to models from OpenAI. However, the startup’s success has caused concern among some regulators in the United States, who believe the company’s technology could present a potential national security risk.
Regardless, DeepSeek is continuing to develop its AI platform. The open MIT license allows developers and businesses to freely experiment with and integrate R1 into their products, although the model demands considerable processing power to function.
Source: