How much VRAM does Z-Image Turbo need?
Z-Image Turbo needs about 7.0 GB at 8-bit, or 4.5 GB as a 4-bit GGUF, at its default output size.
- Type
- Image model
- Parameters
- 6.2B
- Text encoder
- Qwen3-4B
- Released
- Nov 2025
- License
- Apache 2.0
Memory at each precision
| Precision | Weights | Peak at 1024 × 1024 |
|---|---|---|
| 16-bit BF16 | 11.5 GB | 12.5 GB |
| 8-bit FP8 · Q8_0 | 5.9 GB | 7.0 GB |
| 6-bit GGUF Q6_K | 4.7 GB | 5.8 GB |
| 5-bit GGUF Q5_K_M | 4.1 GB | 5.1 GB |
| 4-bit GGUF Q4_K_M · NVFP4 | 3.4 GB | 4.5 GB |
The Qwen3-4B text encoder (3.7 GB at FP8) runs first and then waits in system RAM.
What runs it
| Machine | Best fit | Per image |
|---|---|---|
| RTX 3060 12 GB | 8-bit | 36 s |
| RTX 4060 Ti 16 GB | 16-bit | 21 s |
| RTX 3090 | 16-bit | 13 s |
| RTX 4090 | 16-bit | 6.4 s |
| RTX 5090 | 16-bit | 5.2 s |
| RTX PRO 6000 Blackwell | 16-bit | 2.8 s |
| M4 Max 128 GB | 16-bit | 59 s |
| M5 Max 128 GB | 16-bit | 18 s |
| Ryzen AI Max+ 395 128 GB | 16-bit | 27 s |
| DGX Spark 128 GB | 16-bit | 15 s |
| M3 Ultra 256 GB | 16-bit | 33 s |
| M5 Ultra 512 GB | 16-bit | 11 s |
| H100 SXM | 16-bit | 1.9 s |
| DGX Station GB300 | 16-bit | 1.4 s |
1024 × 1024, 8 steps. Times are rough, plus or minus 30%.
Where to get it
Questions
How much VRAM does Z-Image Turbo need?
Z-Image Turbo needs about 7.0 GB at 8-bit, or 4.5 GB as a 4-bit GGUF, at its default output size.
Can Z-Image Turbo run on an RTX 4090?
Yes. At 16-bit it takes about 6.4 s per image.