Model comparison

MiniMax-H3 vs Wan2.2-Fun-A14B-Control

At Q4_K_M, Wan2.2-Fun-A14B-Control is the smaller download — 9,663,556,352 bytes against 14,576,977,960.

From the file· summed bytes, KV per layer

Side by side

MiniMax-H3Wan2.2-Fun-A14B-Control
Parameters33.1B14.3B
Architectureqwen3vlwan
Layers
Native context
Mixture of expertsnono
Quantizations published2226
Smallest quantization10.60 GiB4.95 GiB
Q4_K_M13.58 GiB9.00 GiB
Licenceotherapache-2.0

KV cache by context

the term that decides long-context viability
ContextMiniMax-H3Wan2.2-Fun-A14B-ControlRatio
4,096
8,192
16,384
32,768
65,536
131,072