Model comparison

granite-4.0-h-350m 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

granite-4.0-h-350mced-base
Parameters340M86M
Architecturegranitehybridced
Layers32
Native context32,768
Mixture of expertsnono
Quantizations published293
Smallest quantization0.15 GiB0.12 GiB
Q4_K_M0.21 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextgranite-4.0-h-350mced-baseRatio
4,0960.02 GiB
8,1920.03 GiB
16,3840.06 GiB
32,7680.13 GiB
65,5360.25 GiB
131,0720.50 GiB