News for hackers. Stuff that matters.

A comment on Production RAG: what I learned from processing 5M+ documents.

Re:Production RAG: what I learned from processing 5M+ documents by 383toast on Monday October 20, 2025 @01:39PM (#45646734)
They should've tested other embedding models, there are better ones than openai's (and cheaper)
[ Parent | on HN ]
Re:Production RAG: what I learned from processing 5M+ documents by prettyblocks on Monday October 20, 2025 @01:44PM (#45646823)
Which do you suggest?
[ Parent | on HN ]
Re:Production RAG: what I learned from processing 5M+ documents by roze_sha on Monday October 20, 2025 @01:57PM (#45646987)
https://huggingface.co/spaces/mteb/leaderboard
[ Parent | on HN ]
Re:Production RAG: what I learned from processing 5M+ documents by 383toast on Monday October 20, 2025 @02:06PM (#45647106)
yep
[ Parent | on HN ]
Re:Production RAG: what I learned from processing 5M+ documents by remz14 on Monday October 20, 2025 @04:45PM (#45649118)
You should use RTEB instead. See here for why: https://huggingface.co/blog/rteb

Here is that leaderboard https://huggingface.co/spaces/mteb/leaderboard?benchmark_nam...

Voyage-3-large seems like SOTA right now

[ Parent | on HN ]
Re:Production RAG: what I learned from processing 5M+ documents by leftnode on Monday October 20, 2025 @03:10PM (#45647899)
The Qwen3 600M and 4B embedding models are near state of the art and aren't too computationally intensive.
[ Parent | on HN ]