Model comparison

Fun-CosyVoice3-0.5B-2512 vs OmniVoice

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

From the file· summed bytes, KV per layer

Side by side

Fun-CosyVoice3-0.5B-2512OmniVoice
Parameters613M
Architecturecosyvoice3-campplusomnivoice-lm
Layers
Native context
Mixture of expertsnono
Quantizations published37
Smallest quantization0.34 GiB0.56 GiB
Q4_K_M0.61 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextFun-CosyVoice3-0.5B-2512OmniVoiceRatio
4,096
8,192
16,384
32,768
65,536
131,072