Model comparison

bitnet-b1.58-2B-4T 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

bitnet-b1.58-2B-4Tced-base
Parameters850M86M
Architectureced
Layers30
Native context4,096
Mixture of expertsnono
Quantizations published23
Smallest quantization0.10 GiB0.12 GiB
Q4_K_M
Licencemitapache-2.0

KV cache by context

the term that decides long-context viability
Contextbitnet-b1.58-2B-4Tced-baseRatio
4,0960.29 GiB
8,1920.59 GiB
16,3841.17 GiB
32,7682.34 GiB
65,5364.69 GiB
131,0729.38 GiB