18 ms·"yet"by nextworddev 3y ago"yet"coding123 3y agoIt's also much slower. LLMs are generating text token at a time. That's not very good for search. Pre-search tokenization however, probably a good fit for LLMs.