Model comparison

gemma-4-12B-it-qat-q4_0-unquantized vs Huihui-Qwen3-Coder-Next-abliterated

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: Huihui-Qwen3-Coder-Next-abliterated's KV cache at 32K is 3.3× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

gemma-4-12B-it-qat-q4_0-unquantizedHuihui-Qwen3-Coder-Next-abliterated
Parameters12.0B79.7B
Architecturegemma4qwen3next
Layers4848
Native context262,144262,144
Mixture of expertsnoyes, 512 experts
Quantizations published273
Smallest quantization6.50 GiB15.07 GiB
Q4_K_M—45.22 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextgemma-4-12B-it-qat-q4_0-unquantizedHuihui-Qwen3-Coder-Next-abliteratedRatio
4,0960.72 GiB0.09 GiB7.66×
8,1920.97 GiB0.19 GiB5.16×
16,3841.47 GiB0.38 GiB3.92×
32,7682.47 GiB0.75 GiB3.29×
65,5364.47 GiB1.50 GiB2.98×
131,0728.47 GiB3.00 GiB2.82×