Chinese AI startup DeepSeek has rolled out an updated version of its R1 reasoning model on the popular developer platform Hugging Face. The release was quietly announced via a WeChat post and marks a continuation of DeepSeek's effort to challenge leading players in the AI field, including US-based OpenAI.

The new R1 model is a minor update, released under the permissive MIT license, allowing commercial use. However, the Hugging Face repository only includes the model's configuration files and internal weights, with no public documentation describing the model's capabilities.

These 'weights' - the parameters that determine how an AI model behaves - total a massive 685 billion, making the model too large to run on standard consumer hardware without special adaptations. DeepSeek gained attention earlier this year when the original R1 demonstrated competitive performance against leading models from OpenAI.

That advancement hasn't gone unnoticed by international observers; US regulators have expressed concern over the potential national security risks posed by DeepSeek's technologies. Despite the controversy, the company continues to make bold moves on the global AI stage.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!