AI

DeepSeek updates its R1 reasoning AI model, releases it on Hugging Face

Chinese Startup Deepseek has an updated version of its R1 -Reeding AI model on the Developer platform Cuddling Face After it announced on Wednesday morning in a WeChat message.

The updated R1, which is under a tolerant mit license, which means that it can be used commercially, is a “small” upgrade, according to the announcement of deep chat. The hugging facial repository contains no description of the model – only configuration files and weights, the internal components of a model that guides the behavior.

With a weight of 685 billion parameters in size, the updated R1 is fairly large. (“Parameters” is synonymous with “weights.”) Without adjustment, the model cannot be performed on hardware of consumer quality.

Deepseek was released earlier this year after the release of R1, which gave models of OpenAi a run for their money. The startup has raised the anger of some supervisors, who claim that the technology of Deepseek is a national security risk.

Source link

See also  How Single Tokens Can Make or Break AI Reasoning

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button