Z-Image-Turbo — OpenVINO INT4
OpenVINO IR (INT4) conversion of Tongyi-MAI/Z-Image-Turbo (6B single-stream DiT, 8-step).
Why OpenVINO?
- Runs on any CPU with no CUDA/NVIDIA dependency, plus GPU acceleration where available (AMD, Nvidia, Intel)
- INT4 weights fit 8GB VRAM cards (transformer
6.2GB→3.1GB); 8-step Turbo stays sub-second - Smaller downloads vs BF16 checkpoints
Conversion
| Parameter | Value |
|---|---|
| Tool | optimum-cli export openvino --weight-format int4 (device-neutral, no vendor passes) |
| Precision | INT4 weights (group-size 64) |
| Source | Tongyi-MAI/Z-Image-Turbo (Apache-2.0) |
Usage (any device)
import openvino as ov
core = ov.Core() # auto-picks CPU/GPU available on your machine
model = core.read_model("model.xml")
compiled = core.compile_model(model, "AUTO")