Model comparison

SenseVoiceSmall vs GigaAM-v3

At Q4_K_M, SenseVoiceSmall is the smaller download — 145,738,304 bytes against 182,497,888.

From the file· summed bytes, KV per layer

Side by side

SenseVoiceSmallGigaAM-v3
Parameters234M223M
Architecturesensevoicegigaam
Layers
Native context
Mixture of expertsnono
Quantizations published918
Smallest quantization0.13 GiB0.17 GiB
Q4_K_M0.14 GiB0.17 GiB
Licenceothermit

KV cache by context

the term that decides long-context viability
ContextSenseVoiceSmallGigaAM-v3Ratio
4,096
8,192
16,384
32,768
65,536
131,072