Model comparison

medgemma-27b-it vs Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP

At Q4_K_M, medgemma-27b-it is the smaller download — 16,546,689,536 bytes against 36,545,830,912. At long context the gap widens: Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP's KV cache at 32K is 1.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

medgemma-27b-itQwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP
Parameters28.8B27.8B
Architecturegemma3qwen35
Layers6264
Native context131,072262,144
Mixture of expertsnono
Quantizations published6834
Smallest quantization5.83 GiB10.12 GiB
Q4_K_M15.41 GiB34.04 GiB
Licenceotherapache-2.0

KV cache by context

the term that decides long-context viability
Contextmedgemma-27b-itQwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPRatio
4,0960.92 GiB0.25 GiB3.68×
8,1921.23 GiB0.50 GiB2.47×
16,3841.86 GiB1.00 GiB1.86×
32,7683.11 GiB2.00 GiB1.55×
65,5365.61 GiB4.00 GiB1.40×
131,07210.61 GiB8.00 GiB1.33×