Hacker News new | ask | show | jobs
by meepmorp 997 days ago
I've seen 1.5 chars/token used as a rule of thumb for estimating token counts in Chinese text.