Model comparison

Wan-Dancer-14B vs MiniMax-H3

At Q4_K_M, MiniMax-H3 is the smaller download — 14,576,977,960 bytes against 26,083,251,744.

From the file· summed bytes, KV per layer

Side by side

Wan-Dancer-14BMiniMax-H3
Parameters17.2B33.1B
Architecturewanqwen3vl
Layers
Native context
Mixture of expertsnono
Quantizations published522
Smallest quantization15.83 GiB10.60 GiB
Q4_K_M24.29 GiB13.58 GiB
Licenceapache-2.0other

KV cache by context

the term that decides long-context viability
ContextWan-Dancer-14BMiniMax-H3Ratio
4,096
8,192
16,384
32,768
65,536
131,072