Z-Image-Turbo ยท FP8 (transformer)
FP8 (float8_e4m3fn) quantization of the Z-Image DiT (distilled turbo) transformer from Tongyi-MAI/Z-Image-Turbo. Quantized by axteam.
| File | Size |
|---|---|
Z-Image-Turbo-transformer-fp8_e4m3fn.safetensors |
6.2 GB |
What this is
- Only the transformer (DiT) is quantized to
float8_e4m3fn(2-D weight tensors; norms and biases stored in BF16, converted from the original FP32). About 25% of the size of the original FP32 transformer (24.6 GB), or half the size of a BF16 copy. - VAE and text encoder(s) are NOT included โ use the ones from the original
Tongyi-MAI/Z-Image-Turbo.
Checked on GPU (diffusers)
On 2026-11-02 the FP8 values of this file (the same values ComfyUI loads; upcast to BF16 for this run) were loaded into the transformer of the original Tongyi-MAI/Z-Image-Turbo pipeline (diffusers 0.42.1, NVIDIA B200; all tensor names matched the original transformer โ 0 missing, 0 unexpected) and compared with the original pipeline run in BF16 at the same prompt and seed (1024ร1024, 9 steps, guidance 0.0).
| Prompt | Pixel PSNR vs original (BF16 run) | Mean abs. pixel diff |
|---|---|---|
| A red apple on a wooden table, studio lighting, photorealistic | 37.4 dB | 2.4 / 255 |
| A cozy cafe street in Seoul at night with warm lights, illustration | 19.2 dB | 13.7 / 255 |
Both images match the BF16 ones closely, with the model's recommended Turbo settings: the apple is practically identical and the street scene keeps the same composition, with small differences in details such as the sign lettering; clean, normal quality. Pixel PSNR is strict for diffusion models: a small weight change shifts the sampling path, so a detailed scene can change its layout while the subject stays correct.
Requirements
| Runtime | ComfyUI (Load Diffusion Model node, float8_e4m3fn single file). Not run in ComfyUI here (no ComfyUI on this server); the FP8 weights were checked with diffusers โ see Checked on GPU above. |
| Also needed | the Qwen3 text encoder (8.0 GB) and VAE from the base repo. |
| Memory | this file is 6.2 GB; add the text encoder(s) and VAE above; the base model card says it fits in 16 GB VRAM. |
Usage (ComfyUI)
- Download the FP8 file into ComfyUI's diffusion-model folder:
hf download axteam/Z-Image-Turbo-FP8 Z-Image-Turbo-transformer-fp8_e4m3fn.safetensors --local-dir ComfyUI/models/diffusion_models
- Get the VAE and text encoder(s) from the original
Tongyi-MAI/Z-Image-Turborepo (or the matching ComfyUI repackaged files) and put them inComfyUI/models/vae/andComfyUI/models/text_encoders/. - In your workflow, load this file with the Load Diffusion Model node (
UNETLoader) โ leaveweight_dtypeatdefault(the file is alreadyfloat8_e4m3fn) โ and wire the original VAE / text-encoder loaders as in the base model's reference workflow. - Sampler settings (steps, CFG/guidance, resolution) are the same as for the original model.
The FP8 file is a plain cast of the transformer weights (no scaling), the common ComfyUI
fp8_e4m3fnconvention; norms and biases are stored in BF16 (the original is FP32).
License
Quantized derivative of Tongyi-MAI/Z-Image-Turbo. Original weights belong to the upstream authors; the base model's license applies to this derivative.
ํ๊ตญ์ด ์์ฝ
Tongyi-MAI/Z-Image-Turbo์ transformer(DiT) FP8 ์์ํ ๋น๋ (ComfyUI์ฉ ๋จ์ผ ํ์ผ) ยท ์ ์ axteam
๊ฐ์
| ํญ๋ชฉ | ๋ด์ฉ |
|---|---|
| ์๋ณธ ๋ชจ๋ธ | Tongyi-MAI/Z-Image-Turbo |
| ํ์ | float8_e4m3fn ๋จ์ผ safetensors (ComfyUI์์ ํํ ์ฐ๋ FP8 ํ์) |
| ํ์ผ | Z-Image-Turbo-transformer-fp8_e4m3fn.safetensors 6.2 GB |
| ํฌํจ | transformer(DiT) |
์์ํ ์์ธ
- ์์ํ ์ ์ฐจ: transformer์ 2D ๊ฐ์ค์น๋ฅผ ์ค์ผ์ผ ์์ด
float8_e4m3fn์ผ๋ก ๋จ์ ํ๋ณํ(cast) โ ComfyUI์์ ํํ ์ฐ๋ ๋ฐฉ์ - BF16 ์ ์ฅ(์๋ณธ FP32์์ ๋ณํ): ์ ๊ทํ์ธต, bias
- ํฌ๊ธฐ: ์๋ณธ FP32 transformer(24.6 GB)์ ์ฝ 25% (BF16 ์ฌ๋ณธ์ ์ ๋ฐ)
- ๋ฏธํฌํจ: VAEยทํ ์คํธ ์ธ์ฝ๋ โ ์๋ณธ ๋ฆฌํฌ์์ ๋ฐ์ ์ฌ์ฉ
- ํ์ง(์ค์ธก, 2026-11-02): diffusers 0.42.1๋ก ์๋ณธ ํ์ดํ๋ผ์ธ์ BF16์ผ๋ก ์คํํ ๊ทธ๋ฆผ๊ณผ, ๊ฐ์ ํ์ดํ๋ผ์ธ์ transformer์ ์ด ํ์ผ์ FP8 ๊ฐ(ComfyUI๊ฐ ์ฝ๋ ๊ฒ๊ณผ ๊ฐ์ ๊ฐ, ์ด ์ ๊ฒ์์๋ BF16์ผ๋ก ์ฌ๋ ค ๊ณ์ฐ)์ ๋ฃ์ด ๊ทธ๋ฆฐ ๊ทธ๋ฆผ์ ๊ฐ์ ํ๋กฌํํธยท์๋ 2๊ฐ๋ก ๋น๊ต(1024ร1024) โ Turbo ๊ถ์ฅ ์ค์ ์์ ๋ ๊ทธ๋ฆผ ๋ชจ๋ ์๋ณธ๊ณผ ๊ฐ๊น์, ์ฌ๊ณผ๋ ์ฌ์ค์ ๊ฐ๊ณ ๊ฑฐ๋ฆฌ ์ฅ๋ฉด๋ ๊ฐ์ ๊ตฌ๋(๊ฐํ ๊ธ์ ๋ฑ ์ธ๋ถ๋ง ๋ค๋ฆ), ์ ์ ํ์ง, ํฝ์ PSNR 37.4ยท19.2 dB (๋ํจ์ ์ ๊ฐ์ค์น๊ฐ ์กฐ๊ธ๋ง ๋ฌ๋ผ๋ ์์ฑ ๊ฒฝ๋ก๊ฐ ๋ฐ๋์ด ๋ณต์กํ ์ฅ๋ฉด์ ๊ตฌ๋๊ฐ ๋ฌ๋ผ์ง ์ ์์, ์ ์์ด ๋ณธ๋ฌธ ํ ์ฐธ๊ณ )
์๊ตฌ์ฌํญ
- ์คํ ํ๊ฒฝ: ComfyUI โ Load Diffusion Model ๋
ธ๋๋ก
float8_e4m3fn๋จ์ผ ํ์ผ ๋ก๋ (ComfyUI๋ก๋ ์คํ ๋ฏธํ์ธ โ FP8 ๊ฐ์ค์น๋ diffusers๋ก ์์ฑ ํ์ธ, ํ์ง ํญ๋ชฉ ์ฐธ๊ณ ) - ์ถ๊ฐ๋ก ํ์ํ ํ์ผ: ์๋ณธ ๋ฆฌํฌ์ Qwen3 ํ ์คํธ ์ธ์ฝ๋(8.0 GB)์ VAE
- ๋ฉ๋ชจ๋ฆฌ: ์ด ํ์ผ 6.2 GB + ์ ํ ์คํธ ์ธ์ฝ๋ยทVAE โ ์๋ณธ ์นด๋ ๊ธฐ์ค 16 GB VRAM์์ ๋์
์ฌ์ฉ๋ฒ (ComfyUI)
- ํ์ผ ๋ฐ๊ธฐ:
hf download axteam/Z-Image-Turbo-FP8 Z-Image-Turbo-transformer-fp8_e4m3fn.safetensors --local-dir ComfyUI/models/diffusion_models - ํ
์คํธ ์ธ์ฝ๋ยทVAE: ์๋ณธ ๋ฆฌํฌ์์ ๋ฐ์
ComfyUI/models/text_encoders/ยทComfyUI/models/vae/์ ๋ฐฐ์น - ๋ถ๋ฌ์ค๊ธฐ: Load Diffusion Model ๋
ธ๋(
UNETLoader),weight_dtype์default(ํ์ผ์ด ์ด๋ฏธfloat8_e4m3fn) - ์ํ๋ฌ ์ค์ : ์คํ ยทCFG(guidance)ยทํด์๋๋ ์๋ณธ ๋ชจ๋ธ๊ณผ ๋์ผ
๋ผ์ด์ ์ค
- ๋ผ์ด์ ์ค: Apache-2.0 โ ์๋ณธ ๋ชจ๋ธ์ ๋ผ์ด์ ์ค๋ฅผ ๊ทธ๋๋ก ๋ฐ๋ฆ
- ๊ถ๋ฆฌ: ์๋ณธ ๊ฐ์ค์น๋ Tongyi-MAI(Alibaba) ์์ , ์ด ๋ฆฌํฌ๋ ์์ํ ํ์๋ฌผ
Model tree for axteam/Z-Image-Turbo-FP8
Base model
Tongyi-MAI/Z-Image-Turbo