Deepseek updates the R1 logical AI model, and it is fired on embrace

Deepseek has released the Chinese starting up an updated copy of Artificial intelligence model R1 logic on The platform for developers embraces the face After announcing it in the WeChat Message on Wednesday morning.

The updated R1, which is subject to the Massachusetts Institute of Technology, which means that it can be used commercially, is a “secondary” upgrade, according to the WeChat announcement from Deepseek. The embraced facial warehouse does not have a description of the model – only the composition files and weights, the internal components of the form that directs its behavior.

It weighs 685 billion in size, the updated R1 is very huge. (“The parameters” synonymous with “weights”.) Without modification, the model is likely to be unable to run on devices at the consumer level.

Dibsic rose to the forefront earlier this year After the R1 version, which gave models from Openai operation for their money. Startup Anger Some organizers in the United States, who argue that Depsic technology is a threat to national security.