Model comparison

ideogram-4-fp8 vs Z-Image-Turbo

At Q4_K_M, Z-Image-Turbo is the smaller download — 4,981,532,736 bytes against 5,027,785,568.

From the file· summed bytes, KV per layer

Side by side

ideogram-4-fp8Z-Image-Turbo
Parameters9.3B6.2B
Architectureqwen3vllumina2
Layers
Native context
Mixture of expertsnono
Quantizations published1930
Smallest quantization4.46 GiB2.41 GiB
Q4_K_M4.68 GiB4.64 GiB
Licenceotherapache-2.0

KV cache by context

the term that decides long-context viability
Contextideogram-4-fp8Z-Image-TurboRatio
4,096
8,192
16,384
32,768
65,536
131,072
ideogram-4-fp8 vs Z-Image-Turbo — size, memory and hardware fit — ossmodeldb