Model comparison

Qwopus3.5-9B-v3.5 vs gemma-4-31B-it

At Q4_K_M, Qwopus3.5-9B-v3.5 is the smaller download — 5,629,104,928 bytes against 18,323,733,440. At long context the gap widens: Qwopus3.5-9B-v3.5's KV cache at 32K is 6.2× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Qwopus3.5-9B-v3.5gemma-4-31B-it
Parameters9.7B31.3B
Architectureclipgemma4
Layers3260
Native context262,144262,144
Mixture of expertsnono
Quantizations published2255
Smallest quantization3.56 GiB7.95 GiB
Q4_K_M5.24 GiB17.07 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwopus3.5-9B-v3.5gemma-4-31B-itRatio
4,0960.13 GiB1.80 GiB14.36×
8,1920.25 GiB2.42 GiB9.68×
16,3840.50 GiB3.67 GiB7.34×
32,7681.00 GiB6.17 GiB6.17×
65,5362.00 GiB11.17 GiB5.59×
131,0724.00 GiB21.17 GiB5.29×