These Mini AI Models Match OpenAI With 1,000 Times Less Data. - eviltoast
  • General_Effort@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 month ago

    After a quick skim, seems like the article has lots of errors. Molmo is trained on top of Qwen. The smallest ones are trained on something by the same company as Molmo.