Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs Qwen3.6-34B-80L-Fable-5-Heretic

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

gemma-4-12B-it-qat-q4_0-unquantizedQwen3.6-34B-80L-Fable-5-Heretic
Parameters12.0B33.4B
Architecturegemma4qwen35
Layers4880
Native context262,144262,144
Mixture of expertsnono
Quantizations published225
Smallest quantization6.50 GiB12.27 GiB
Q4_K_M19.10 GiB
Licenceapache-2.0

KV cache by context

the term that decides long-context viability
Contextgemma-4-12B-it-qat-q4_0-unquantizedQwen3.6-34B-80L-Fable-5-HereticRatio
4,0960.72 GiB0.31 GiB2.30×
8,1920.97 GiB0.63 GiB1.55×
16,3841.47 GiB1.25 GiB1.17×
32,7682.47 GiB2.50 GiB1.01×
65,5364.47 GiB5.00 GiB1.12×
131,0728.47 GiB10.00 GiB1.18×