Model comparison

Krea-2-Raw vs FLUX.2-klein-9B

At Q4_K_M, FLUX.2-klein-9B is the smaller download — 5,027,783,648 bytes against 7,486,289,184.

From the file· summed bytes, KV per layer

Side by side

Krea-2-RawFLUX.2-klein-9B
Parameters12.8B9.1B
Architecturekrea2flux
Layers
Native context
Mixture of expertsnono
Quantizations published2929
Smallest quantization2.34 GiB3.71 GiB
Q4_K_M6.97 GiB4.68 GiB
Licenceotherother

KV cache by context

the term that decides long-context viability
ContextKrea-2-RawFLUX.2-klein-9BRatio
4,096
8,192
16,384
32,768
65,536
131,072