Model comparison

FLUX.2-klein-9b-kv vs FLUX.1-dev

At Q4_K_M, FLUX.2-klein-9b-kv is the smaller download — 5,716,891,936 bytes against 6,931,817,760.

From the file· summed bytes, KV per layer

Side by side

FLUX.2-klein-9b-kvFLUX.1-dev
Parameters9.1B11.9B
Architecturefluxflux
Layers
Native context
Mixture of expertsnono
Quantizations published937
Smallest quantization3.33 GiB3.75 GiB
Q4_K_M5.32 GiB6.46 GiB
Licenceotherother

KV cache by context

the term that decides long-context viability
ContextFLUX.2-klein-9b-kvFLUX.1-devRatio
4,096
8,192
16,384
32,768
65,536
131,072
FLUX.2-klein-9b-kv vs FLUX.1-dev — size, memory and hardware fit — ossmodeldb