Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Did Google actually train a new model? The cutoff dates for Gemini 3 and 2.5 are the same.




I think this simply suggests the same (or very similar) training corpora.

Surely, they would throw current events, news articles, the latest snapshot of WikiPedia, etc...

I can't imagine it making sense to purposefully neglect to keep a model as up-to-date as possible!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: