← sources

Qwen 3.8 27B Q5 vs Qwen 3.8 Next Q3_S for document analysis

source: reddit (reddit:LocalLLaMA)published 2026-10-11 06:52ingested 2026-10-12 02:39

retained text

I have an app that processes \~500 different documents every day. An agent analyzes them, sorts them, tags them, extracts various legal nuances, creates descriptions, etc.

So far I've been using Qwen 3.8 27B Q5 with a 150K context window, but I'm wondering whether I should switch to Qwen 3.8 Next Q3\_S, since it has much more knowledge and could extract data much better than the 27B.

However, I'm hesitant because it's a MoE model and Q3 is quite a low quantization, so I'm worried the results could end up being worse.

Everything  run on 64 GB of RAM and an R9700.

document fetched 2026-10-12 00:37 · canonical https://reddit.com/r/LocalLLaMA/comments/1x31ak8/qwen_38_27b_q5_vs_qwen_38_next_q3_s_for_document