Model comparison

MiniMax-H3 vs lingbot-video-moe-30b-a3b

At Q4_K_M, MiniMax-H3 is the smaller download — 14,576,977,960 bytes against 17,026,985,056.

From the file· summed bytes, KV per layer

Side by side

MiniMax-H3lingbot-video-moe-30b-a3b
Parameters33.1B30.1B
Architectureqwen3vlwan
Layers
Native context
Mixture of expertsnono
Quantizations published222
Smallest quantization10.60 GiB12.12 GiB
Q4_K_M13.58 GiB15.86 GiB
Licenceotherother

KV cache by context

the term that decides long-context viability
ContextMiniMax-H3lingbot-video-moe-30b-a3bRatio
4,096
8,192
16,384
32,768
65,536
131,072
MiniMax-H3 vs lingbot-video-moe-30b-a3b — size, memory and hardware fit — ossmodeldb