Skip to main content

"Fast" DeepSeek-R1

A fast variant of the model would be greatly appreciated. The base model is a bit too slow since there’re lots of output tokens.

4 comments

Log in to comment and vote

Comments4

  • mhr305

    •

    Mar 7, 2025

    both fast and base for R1 and distilled r1 llama 70B are available :

    https://studio.nebius.com/

  • Wilame Souza

    •

    Feb 4, 2025

    Would you pay if the fast version was almost twice as expensive? If you compare the base version and the fast version of the LlaMa 3.3 you will see that the price difference is almost double. The DeepSeek R1 has a lot of power and uses a lot of resources, and I'm sure a fast version would be so expensive that it would be a "luxury item". I think this is the same reason why they don't have a LlaMa 3.1 405b fast