`UD-Q4_K_XL` or `Q4_K_M`?
- what does
UDmean in the name? - why Q4_K_XL is smaller than Q4_K_M? I think XL is supposed to be larger than M.
- what to choose?
wondering the same
Pretty sure UD means unsloth dynamic, so based on their blog posts, you'll want that.
Why would they continue to upload the old quants when their UD quants are significantly better?
Q4_K_XL decides to use Q5_K on important matrices if it considers it safe. Q4_K_M uses Q6_K there mostly.
Most matrices don't differ and use Q4_K.
I'd always go with the XL variant.
For better results, Always use the XL quants.
yes, UD means Unsloth Dynamic
For better results, Always use the XL quants.
yes, UD means Unsloth Dynamic
So no UD means not using it? But you said all Qwen3 models are using it, right?
If the highest quant/size i can run is Q3_K_M.gguf, then does it make sense to download UD-Q3_K_XL.gguf instead? That's all i want to know :)
(because its also smaller in size, but performance?)