Reviewing Google's extensive LLM training pre-ChatGPT

soldni · x · 2026-08-28

Commenting on a table of numerous LLMs trained by Google before ChatGPT, the author notes that DeepMind and Google Brain were separate entities then, making double the output expected. They also acknowledge some overcounting in the table (e.g., Minerva, FlanT5 were relatively cheap fine-tunes), but affirm how early they started.

Related event: Google Had Already Trained Many LLMs Before ChatGPT(2 posts)→

Original post →

More from Companies & People

Companies & People channel →