Model comparison

MiniMax-H3 vs Wan2.1_T2V_14B_FusionX_VACE

At Q4_K_M, Wan2.1_T2V_14B_FusionX_VACE is the smaller download — 11,629,623,072 bytes against 14,576,977,960.

From the file· summed bytes, KV per layer

Side by side

MiniMax-H3Wan2.1_T2V_14B_FusionX_VACE
Parameters33.1B17.3B
Architectureqwen3vlwan
Layers
Native context
Mixture of expertsnono
Quantizations published2215
Smallest quantization10.60 GiB5.92 GiB
Q4_K_M13.58 GiB10.83 GiB
Licenceotherapache-2.0

KV cache by context

the term that decides long-context viability
ContextMiniMax-H3Wan2.1_T2V_14B_FusionX_VACERatio
4,096
8,192
16,384
32,768
65,536
131,072
MiniMax-H3 vs Wan2.1_T2V_14B_FusionX_VACE — size, memory and hardware fit — ossmodeldb