Model comparison

L3.2-Rogue-Creative-Instruct-Uncensored-Abliterated-7B vs gemma-4-12B-it-qat-q4_0-unquantized

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-12B-it-qat-q4_0-unquantized's KV cache at 32K is 3.4× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

L3.2-Rogue-Creative-Instruct-Uncensored-Abliterated-7Bgemma-4-12B-it-qat-q4_0-unquantized
Parameters7.5B12.0B
Architecturellamagemma4
Layers6748
Native context131,072262,144
Mixture of expertsnono
Quantizations published102
Smallest quantization2.73 GiB6.50 GiB
Q4_K_M4.27 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextL3.2-Rogue-Creative-Instruct-Uncensored-Abliterated-7Bgemma-4-12B-it-qat-q4_0-unquantizedRatio
4,0961.05 GiB0.72 GiB1.46×
8,1922.09 GiB0.97 GiB2.16×
16,3844.19 GiB1.47 GiB2.85×
32,7688.38 GiB2.47 GiB3.39×
65,53616.75 GiB4.47 GiB3.75×
131,07233.50 GiB8.47 GiB3.96×