logoalt Hacker News

cristoperbtoday at 4:22 AM1 replyview on HN

I haven't tried it for anything myself yet. The paper provides several benchmarks. The emphasis during training was on multi-language support (over 1800 languages are represented in its pre-training data, which is 40% non-English) and non-copyrighted training data... and the benchmarks seem to suffer for it.

https://arxiv.org/abs/2509.14233


Replies

nicolarictoday at 5:46 AM

it's quite bad tbh. i've tried it for some time and i expected much more...