A 30-game battle royale across eleven LLMs, $482 of inference, and one finding that should change how you read model benchmarks.
  • SmokeInFog@midwest.socialEnglish
    4·
    il y a 2 mois

    I feel like this article is a bunch of non-sequiturs strung together. He even claims that his LLM agent battle is something “a routing customer actually cares about.”

      • Helix 🧬@feddit.orgEnglish
        2·
        il y a 2 mois

        Yep. The writing style makes me aggressive.

        “There was one who did this, another one did this. The third will surprise you. And it was not this. But this.”

        Jesus Christ why did they train these things on Buzzfeed listicles and corporate blog slop?!??

  • Em Adespoton@lemmy.ca
    2·
    il y a 2 mois

    Either is fine. I’d just send it a wall of text and it would hit its token limit and stall.

    Unless it was running towards me to protect me; in that case, I’d want Claude.