Z-Image-Turbo: Self-Hosting & Deployment Guide
Z-Image-Turbo is an open-weight text-to-image model.
What is Z-Image-Turbo?
Z-Image-Turbo is an open-weight text-to-image model, published by Tongyi-MAI with open weights at Tongyi-MAI/Z-Image-Turbo. LLM API does not route this model today, so there is no endpoint or code sample for it here. What this page does give you is the self-hosting picture: published weight sizes, the VRAM each quantisation needs and the commands to serve it yourself. Every figure is read from the public repository and refreshed weekly. At import it had 629,415 downloads on Hugging Face in the previous 30 days, under the apache-2.0 licence.
How to run Z-Image-Turbo yourself
Open weights (apache-2.0), 6.2B parameters. Below: how much memory the weights need at each quantization, which hardware fits them, and the commands to serve it.
Z-Image-Turbo VRAM requirements
| Precision / quantization | Weights | Figure | Source |
|---|---|---|---|
| BF16 (full precision) | 12.3 GB | Estimate | 6.2B params × 2 bytes |
| GGUF Q8_0 | 7.2 GB | File size | unsloth/Z-Image-Turbo-GGUF |
| GGUF Q4_K_M | 5 GB | File size | unsloth/Z-Image-Turbo-GGUF |
File size = actual download size on Hugging Face. Estimate = parameter count × bytes per weight. Weights only — the KV cache for long contexts needs extra memory; the fit tables below add 15% headroom as a rule of thumb.
GPU requirements: deploy Z-Image-Turbo on your hardware
| Hardware | BF16 (full precision) | GGUF Q8_0 | GGUF Q4_K_M |
|---|---|---|---|
| NVIDIA H200141 GB | 1× | 1× | 1× |
| NVIDIA H10080 GB | 1× | 1× | 1× |
| NVIDIA A100 80GB80 GB | 1× | 1× | 1× |
| NVIDIA L40S48 GB | 1× | 1× | 1× |
GPUs needed per variant.
| Hardware | GGUF Q8_0 | GGUF Q4_K_M |
|---|---|---|
| RTX 509032 GB | 1× | 1× |
| RTX 409024 GB | 1× | 1× |
| RTX 309024 GB | 1× | 1× |
GGUF via llama.cpp; more cards or CPU offload for larger files.
| Hardware | GGUF Q8_0 | GGUF Q4_K_M |
|---|---|---|
| Mac, 512 GBM3 Ultra | Fits | Fits |
| Mac, 192 GBM2 Ultra | Fits | Fits |
| Mac, 128 GBM4 Max / M3 Max | Fits | Fits |
| Mac, 64 GBM4 Pro / Max | Fits | Fits |
Assumes ~75% of unified memory is usable by the GPU.
| Hardware | BF16 (full precision) | GGUF Q8_0 | GGUF Q4_K_M |
|---|---|---|---|
| AWS p5.48xlarge8× H100 · 640 GB | Fits | Fits | Fits |
| AWS p4de.24xlarge8× A100 80GB · 640 GB | Fits | Fits | Fits |
| AWS g6e.12xlarge4× L40S · 192 GB | Fits | Fits | Fits |
| Google Cloud a3-highgpu-8g8× H100 · 640 GB | Fits | Fits | Fits |
| Azure ND H100 v58× H100 · 640 GB | Fits | Fits | Fits |
Whole-instance GPU memory; public instance specs.
Not on LLM.API yet
We do not route this model through our API at the moment. Browse the models you can call today — most workloads have a close match already live.
Not available on LLM API yet
We do not route this model through our API at the moment, so there is no endpoint or code snippet for it yet. Browse the models you can call today — most workloads have a close match already live.
Browse available modelsWhen to Use — When NOT to Use
Use it if...
- You need images generated from a text prompt, or edited from a reference image (Z-Image-Turbo)
- You want image generation on the same LLM API key as your text models, with no extra vendor SDK
- Your pipeline can take the result as a URL or base64 payload rather than a streamed response
- You want per-image pricing you can read off the catalogue before you commit
Avoid if...
- You need a chat, coding or reasoning model — this model returns images, not text
- You need token-by-token streaming: image calls return once the picture is finished
- You need video, speech or embeddings; those are separate models in the catalogue
- Your licensing review has not cleared generated imagery for commercial use
Get one key to every model
Swap your API key. Keep your code.