Model comparison

s2-pro vs VoxCPM2

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

From the file· summed bytes, KV per layer

Side by side

s2-proVoxCPM2
Parameters4.6B2.3B
Architecturefish-speechvoxcpm-acoustic
Layers36
Native context
Mixture of expertsyes, 1 expertsno
Quantizations published77
Smallest quantization2.40 GiB1.57 GiB
Q4_K_M3.32 GiB
Licenceotherapache-2.0

KV cache by context

the term that decides long-context viability
Contexts2-proVoxCPM2Ratio
4,096
8,192
16,384
32,768
65,536
131,072