That doesn't make any sense, because their 8B is listed as benchmarking above th... | Hacker News

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

lolinder on Nov 30, 2024 | parent | context | favorite | on: DeepThought-8B: A small, capable reasoning model

That doesn't make any sense, because their 8B is listed as benchmarking above the 13B "model A".

sigmoid10 on Dec 2, 2024 [–]

That's why it is very likely it has seen more tokens during training and why the plot is worthless.

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