Model comparison

MiniMax-H3 vs SCAIL-2

At Q4_K_M, SCAIL-2 is the smaller download — 10,862,812,832 bytes against 14,576,977,960.

From the file· summed bytes, KV per layer

Side by side

MiniMax-H3SCAIL-2
Parameters33.1B16.4B
Architectureqwen3vlwan
Layers
Native context
Mixture of expertsnono
Quantizations published2226
Smallest quantization10.60 GiB5.61 GiB
Q4_K_M13.58 GiB10.12 GiB
Licenceothermit

KV cache by context

the term that decides long-context viability
ContextMiniMax-H3SCAIL-2Ratio
4,096
8,192
16,384
32,768
65,536
131,072