Model comparison

VoxCPM-0.5B vs s2-pro

These two publish different quantization sets; the table below has the exact sizes.

From the file· summed bytes, KV per layer

Side by side

VoxCPM-0.5Bs2-pro
Parameters728M4.6B
Architecturellamafish-speech
Layers36
Native context
Mixture of expertsnoyes, 1 experts
Quantizations published47
Smallest quantization2.45 GiB2.40 GiB
Q4_K_M3.32 GiB
Licenceother

KV cache by context

the term that decides long-context viability
ContextVoxCPM-0.5Bs2-proRatio
4,096
8,192
16,384
32,768
65,536
131,072