Model comparison

ERNIE-Image-Turbo vs FLUX.1-dev

At Q4_K_M, ERNIE-Image-Turbo is the smaller download — 5,019,124,416 bytes against 6,931,817,760.

From the file· summed bytes, KV per layer

Side by side

ERNIE-Image-TurboFLUX.1-dev
Parameters8.0B11.9B
Architecturewanflux
Layers
Native context
Mixture of expertsnono
Quantizations published1937
Smallest quantization2.96 GiB3.75 GiB
Q4_K_M4.67 GiB6.46 GiB
Licenceapache-2.0other

KV cache by context

the term that decides long-context viability
ContextERNIE-Image-TurboFLUX.1-devRatio
4,096
8,192
16,384
32,768
65,536
131,072
ERNIE-Image-Turbo vs FLUX.1-dev — size, memory and hardware fit — ossmodeldb