(blog.abdellatif.io)

548 points tifa2up | 5 comments | 20 Oct 25 15:55 UTC | HN request time: 0.293s | source

Show context

383toast ◴[20 Oct 25 17:39 UTC] No.45646734[source]▶

They should've tested other embedding models, there are better ones than openai's (and cheaper)

replies(1): >>45646823 #

1. prettyblocks ◴[20 Oct 25 17:44 UTC] No.45646823[source]▶

Which do you suggest?

2. roze_sha ◴[20 Oct 25 17:57 UTC] No.45646987[source]▶

3. 383toast ◴[20 Oct 25 18:06 UTC] No.45647106[source]▶

yep

4. leftnode ◴[20 Oct 25 19:10 UTC] No.45647899[source]▶

The Qwen3 600M and 4B embedding models are near state of the art and aren't too computationally intensive.

5. remz14 ◴[20 Oct 25 20:45 UTC] No.45649118[source]▶

You should use RTEB instead. See here for why: https://huggingface.co/blog/rteb

Voyage-3-large seems like SOTA right now

Production RAG: what I learned from processing 5M+ documents