Model comparison

SCAIL-2 vs MiniMax-H3

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

SCAIL-2MiniMax-H3
Parameters16.4B33.1B
Architecturewanqwen3vl
Layers
Native context
Mixture of expertsnono
Quantizations published2622
Smallest quantization5.61 GiB10.60 GiB
Q4_K_M10.12 GiB13.58 GiB
Licencemitother

KV cache by context

the term that decides long-context viability
ContextSCAIL-2MiniMax-H3Ratio
4,096
8,192
16,384
32,768
65,536
131,072