← sources

What’s better? A very small quant or a large model distilled into a small one?

source: reddit (reddit:LocalLLaMA)published 2026-09-30 15:59ingested 2026-10-01 02:44

retained text

I see people asking (begging, take it as humor) for 0.000001 bit quants but isn’t a lower quant essentially going to hurt model’s quality and tool calls? 

document fetched 2026-10-01 00:45 · canonical https://reddit.com/r/LocalLLaMA/comments/1wu8qa0/whats_better_a_very_small_quant_or_a_large_model