Actually much improved benchmark scores considering this is supposed to be a “minor” update

I don’t know why they don’t just call this v4.2, why don’t LLM labs know how to do version numbers?

  • humanspiral@lemmy.ca
    link
    fedilink
    English
    arrow-up
    1
    ·
    3 days ago

    Deepseek does continuous updates on their chat interface LLM. Multilingual SWE is no longer a “benchmax” target, but they are killing it on more obscure languages.

    • damnthefilibuster@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      7 days ago

      Run it off one of the vendors who have serious “won’t use your data for training or other shit” terms. I like DeepInfra for this kind of use.