DeepSeek R1 AI Model Update & Release - Hugging Face

DeepSeek Releases Updated R1 Reasoning AI Model

The Chinese AI company DeepSeek has made a new iteration of its R1 reasoning model available on the Hugging Face developer platform. This release followed an announcement disseminated via a WeChat message on Wednesday morning.

According to DeepSeek’s announcement, this update to R1 is considered a “minor” improvement. The Hugging Face repository currently lacks a descriptive overview of the model.

Model Details and Licensing

Instead, the repository contains only the necessary configuration files and the model's weights. These weights are the internal elements that determine the model’s operational characteristics.

The updated R1 model is substantial in size, boasting 685 billion parameters. The term "parameters" is used interchangeably with "weights."

Due to its considerable size, running the model without hardware modifications is unlikely to be feasible on standard consumer-level computing devices.

DeepSeek's Rise and Regulatory Scrutiny

DeepSeek initially gained recognition earlier in the year with the initial release of R1. This model demonstrated competitive performance against those developed by OpenAI.

The startup has attracted attention from some US regulators. Concerns have been raised regarding potential national security implications stemming from DeepSeek’s technological advancements.

These regulators contend that the technology developed by DeepSeek could present a risk to national security interests.