Model comparison

Ministral-3-8B-Reasoning-2512-heretic vs Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP

At Q4_K_M, Ministral-3-8B-Reasoning-2512-heretic is the smaller download — 5,198,385,440 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 2.1× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

Ministral-3-8B-Reasoning-2512-hereticQwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP
Parameters8.9B27.8B
Architecturemistral3qwen35
Layers3464
Native context262,144262,144
Mixture of expertsnono
Quantizations published334
Smallest quantization4.84 GiB10.12 GiB
Q4_K_M4.84 GiB34.04 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextMinistral-3-8B-Reasoning-2512-hereticQwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPRatio
4,0960.53 GiB0.25 GiB2.12×
8,1921.06 GiB0.50 GiB2.12×
16,3842.13 GiB1.00 GiB2.12×
32,7684.25 GiB2.00 GiB2.12×
65,5368.50 GiB4.00 GiB2.12×
131,07217.00 GiB8.00 GiB2.12×