Model comparison

FireRedPunc vs ced-base

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

From the file· summed bytes, KV per layer

Side by side

FireRedPuncced-base
Parameters102M86M
Architecturefireredpuncced
Layers
Native context
Mixture of expertsnono
Quantizations published33
Smallest quantization0.05 GiB0.12 GiB
Q4_K_M
Licenceapache-2.0apache-2.0

KV cache by context

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