DeepSeek, a large-scale model manufacturer under Quantitative Magic Square, has enabled hard disk caching technology.

On August 2, the private sale giant, Quantitative Magic Cube, announced on its official WeChat account that its new organization, 'DeepSeek,' exploring AGI (Artificial General Intelligence), has deployed context hard drive caching technology to cache content that is expected to be reused in a distributed array of hard drives. In the usage scenario of large model APIs, if there is a duplicate input, the duplicate part only needs to be read from the cache without calculation. The cost of DeepSeek for the cached hit part is 0.1 yuan per million tokens.

View Original
  • Reward
  • Comment
  • Share
Comment
No comments