Model comparison

Bernini-R vs MiniMax-H3

At Q4_K_M, Bernini-R is the smaller download — 9,650,100,736 bytes against 14,576,977,960.

From the file· summed bytes, KV per layer

Side by side

Bernini-RMiniMax-H3
Parameters14.3B33.1B
Architecturewanqwen3vl
Layers
Native context
Mixture of expertsnono
Quantizations published3222
Smallest quantization4.94 GiB10.60 GiB
Q4_K_M8.99 GiB13.58 GiB
Licenceapache-2.0other

KV cache by context

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