Model comparison

Wan_2.1_ComfyUI_repackaged vs MiniMax-H3

At Q4_K_M, MiniMax-H3 is the smaller download — 14,576,977,960 bytes against 68,351,109,664.

From the file· summed bytes, KV per layer

Side by side

Wan_2.1_ComfyUI_repackagedMiniMax-H3
Parameters33.1B
Architecturepigqwen3vl
Layers
Native context
Mixture of expertsnono
Quantizations published4822
Smallest quantization4.25 GiB10.60 GiB
Q4_K_M63.66 GiB13.58 GiB
Licenceapache-2.0other

KV cache by context

the term that decides long-context viability
ContextWan_2.1_ComfyUI_repackagedMiniMax-H3Ratio
4,096
8,192
16,384
32,768
65,536
131,072