DeepSeek Releases an Updated R1 Model on Hugging Face
Although DeepSeek described it as a “minor” upgrade, the model is notably large, boasting 685 billion parameters
Chinese AI startup DeepSeek has released a minor update to its flagship R1 reasoning model on the developer platform Hugging Face.
The updated R1 model, which remains under the permissive MIT license, allows for commercial use and redistribution.
Although DeepSeek described it as a “minor” upgrade, the model is notably large, boasting 685 billion parameters.
These parameters—or weights—are the core elements that guide the model's responses and decision-making capabilities.
Due to its sheer size, the updated R1 is not designed for consumer-grade hardware and will require powerful computing resources to operate effectively.
The Hugging Face repository currently provides only configuration files and model weights, without a detailed description of the update’s specific enhancements.
The announcement was made Wednesday via the company’s official WeChat account.
DeepSeek gained significant attention earlier this year when its original R1 model was seen as a serious challenger to offerings from OpenAI.