DavidAU
/

Qwen3-128k-30B-A3B-NEO-MAX-Imatrix-gguf

Text Generation

creative writing

problem solving

Mixture of Experts

mixture of experts

8 experts activated

Model card Files Files and versions

DavidAU commited on May 8

Commit

e1d1512

·

1 Parent(s): 9cd2d62

Update README.md

Files changed (1) hide show

README.md +6 -2

README.md CHANGED Viewed

@@ -101,8 +101,6 @@ IQ2s will be a lot stronger than the IQ1_Ms.
 IQ3s and higher quants will show a very large change in performance, with IQ4_XS/IQ4_NL being the peak for NEO Imatrix effect
 and specific quality at this quant level.
-NOTE: IQ3s will outperform "Q3s" quants, likewise for IQ2s vs Q2s quants.
 Q4s will be high performance.
 Q5s will be very high performance.
@@ -111,6 +109,12 @@ Q6 will be peak performance, but with minimal NEO imatrix effect(s).
 Q8s (specialized) will be well... really, really great.
 SPECIALIZED QUANTS:
 Some quants will have multiple versions:

 IQ3s and higher quants will show a very large change in performance, with IQ4_XS/IQ4_NL being the peak for NEO Imatrix effect
 and specific quality at this quant level.
 Q4s will be high performance.
 Q5s will be very high performance.
 Q8s (specialized) will be well... really, really great.
+NOTES:
+- IQ3s will outperform Q3s quants, likewise for IQ2s vs Q2s quants.
+- IQ4_XS / IQ4_NL will perform at or outperform Q4s.
+- IQ3_S / IQ3_M are very strong and can be used for most tasks.
+- All Imatrix versions of quants will outperform their non-imatrix counterparts.
 SPECIALIZED QUANTS:
 Some quants will have multiple versions: