Z-Image Turbo VRAM requirements
A 6B, 8-step model from Alibaba's Tongyi lab and one of the most downloaded ComfyUI models of the past year. Fast and light.
TypeImage
Parameters6B
Fits entirely from6 GB
8-bit or better from9 GB
01Measured on my RTX 5060 Ti 16 GB
16-bit · 1024×1024
11.0s / image
z_image_turbo_bf16.safetensors
Peak VRAM14.6 GB
Run range11.0–11.0 s
02The files, and how much VRAM each needs
| File | Size | Needed | Min. VRAM | Quality | Source |
|---|---|---|---|---|---|
| 16-bit | 12.3 GB | 14.3 GB | 15 GB | the original weights | Comfy-Org/z_image_turbo → |
| Q8_0 | 7.2 GB | 9.2 GB | 10 GB | practically identical to the original | unsloth/Z-Image-Turbo-GGUF → |
| INT8 | 6.2 GB | 8.2 GB | 9 GB | practically identical to the original | Comfy-Org/z_image_turbo → |
| Q6_K | 5.9 GB | 7.9 GB | 8 GB | very close to the original | unsloth/Z-Image-Turbo-GGUF → |
| Q5_K_M | 5.6 GB | 7.6 GB | 8 GB | close; small differences in fine detail | unsloth/Z-Image-Turbo-GGUF → |
| Q4_K_M | 5.0 GB | 7.0 GB | 8 GB | good; some loss in fine detail and text | unsloth/Z-Image-Turbo-GGUF → |
| Q3_K_M | 4.2 GB | 6.2 GB | 7 GB | noticeable loss of detail | unsloth/Z-Image-Turbo-GGUF → |
| Q2_K | 3.6 GB | 5.6 GB | 6 GB | heavy loss; a last resort | unsloth/Z-Image-Turbo-GGUF → |
“Needed” = file + 1.2 GB working memory + 0.8 GB system reserve.
Free ComfyUI workflow, tested on my RTX 5060 Ti: 16-bit workflow (.json). Drag the file into ComfyUI. All workflows →
03By amount of VRAM
04Best GPU for Z-Image Turbo
The cheapest cards (by launch price) that run it well, and every card sorted by memory: best GPU for Z-Image Turbo → Planning bigger images or longer clips? Open the calculator →
05By graphics card
| GPU | VRAM | Verdict | Best file | Needed |
|---|---|---|---|---|
| Desktop graphics cards | ||||
| RTX 2060 6 GB | 6 GB | Tight | Q2_K | 5.6 GB |
| RTX 3050 6 GB | 6 GB | Tight | Q2_K | 5.6 GB |
| RTX 2070 Super 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 2080 Super 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3050 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3060 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3060 Ti 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3070 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3070 Ti 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 4060 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 4060 Ti 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 5050 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 5060 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 5060 Ti 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 7600 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 9050 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 9060 XT 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| Arc B570 10 GB | 10 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 3080 10 GB | 10 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 2080 Ti 11 GB | 11 GB | Runs well | Q8_0 | 9.2 GB |
| Arc B580 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 2060 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 3060 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 3080 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 3080 Ti 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 4070 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RTX 4070 Super 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RTX 4070 Ti 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RTX 5070 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RX 7700 XT 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RX 9070 GRE 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| Arc A770 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 4060 Ti 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 4070 Ti Super 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 4080 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 4080 Super 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 5060 Ti 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 5070 Ti 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 5080 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 7600 XT 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 7800 XT 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 7900 GRE 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 9060 XT 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 9070 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 9070 XT 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 7900 XT 20 GB | 20 GB | Runs well | 16-bit | 14.3 GB |
| Arc Pro B60 24 GB | 24 GB | Runs well | 16-bit | 14.3 GB |
| RTX 3090 24 GB | 24 GB | Runs well | 16-bit | 14.3 GB |
| RTX 3090 Ti 24 GB | 24 GB | Runs well | 16-bit | 14.3 GB |
| RTX 4090 24 GB | 24 GB | Runs well | 16-bit | 14.3 GB |
| RX 7900 XTX 24 GB | 24 GB | Runs well | 16-bit | 14.3 GB |
| Arc Pro B70 32 GB | 32 GB | Runs well | 16-bit | 14.3 GB |
| RTX 5090 32 GB | 32 GB | Runs well | 16-bit | 14.3 GB |
| Laptop GPUs | ||||
| RTX 3050 Laptop 4 GB | 4 GB | Offload only | Q2_K | 5.6 GB |
| RTX 3050 Ti Laptop 4 GB | 4 GB | Offload only | Q2_K | 5.6 GB |
| RTX 2060 Laptop 6 GB | 6 GB | Tight | Q2_K | 5.6 GB |
| RTX 3050 Laptop 6 GB | 6 GB | Tight | Q2_K | 5.6 GB |
| RTX 3060 Laptop 6 GB | 6 GB | Tight | Q2_K | 5.6 GB |
| RTX 4050 Laptop 6 GB | 6 GB | Tight | Q2_K | 5.6 GB |
| RTX 2070 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 2070 Super Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 2080 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 2080 Super Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3070 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3070 Ti Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 3080 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 4060 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 4070 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 5050 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 5060 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 5070 Laptop 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 7600M 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 7600M XT 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 7600S 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RX 7700S 8 GB | 8 GB | Runs | Q6_K | 7.9 GB |
| RTX 4080 Laptop 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RTX 5070 Laptop 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RTX 5070 Ti Laptop 12 GB | 12 GB | Runs well | INT8 | 8.2 GB |
| RX 7800M 12 GB | 12 GB | Runs well | Q8_0 | 9.2 GB |
| RTX 3080 Laptop 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 3080 Ti Laptop 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 4090 Laptop 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 5080 Laptop 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RX 7900M 16 GB | 16 GB | Runs well | 16-bit | 14.3 GB |
| RTX 5090 Laptop 24 GB | 24 GB | Runs well | 16-bit | 14.3 GB |
| Unified memory | ||||
| Radeon 8060S (Strix Halo) 96 GB | 96 GB | Runs well | 16-bit | 14.3 GB |
| Radeon 8065S (Gorgon Halo) 160 GB | 160 GB | Runs well | 16-bit | 14.3 GB |
| Apple Silicon Macs (by memory) | ||||
| Mac 16 GB | 12.7 GB | Runs well | Q8_0 | 9.2 GB |
| Mac 18 GB | 14.4 GB | Runs well | 16-bit | 14.3 GB |
| Mac 24 GB | 19.6 GB | Runs well | 16-bit | 14.3 GB |
| Mac 32 GB | 26.8 GB | Runs well | 16-bit | 14.3 GB |
| Mac 36 GB | 30.2 GB | Runs well | 16-bit | 14.3 GB |
| Mac 48 GB | 40.2 GB | Runs well | 16-bit | 14.3 GB |
| Mac 64 GB | 55.7 GB | Runs well | 16-bit | 14.3 GB |
| Mac 96 GB | 85 GB | Runs well | 16-bit | 14.3 GB |
| Mac 128 GB | 115.4 GB | Runs well | 16-bit | 14.3 GB |
| Mac 192 GB | 175.4 GB | Runs well | 16-bit | 14.3 GB |
| Mac 256 GB | 236.9 GB | Runs well | 16-bit | 14.3 GB |
| Mac 512 GB | 498.1 GB | Runs well | 16-bit | 14.3 GB |
06Text encoder, VAE and other files
| File | Folder | Size | When | |
|---|---|---|---|---|
| Qwen3 4B FP8 qwen_3_4b_fp8_mixed.safetensors | models/text_encoders | 5.6 GB | text encoder · smaller, recommended | Download → |
| Qwen3 4B BF16 qwen_3_4b.safetensors | models/text_encoders | 8.0 GB | text encoder · alternative | Download → |
| FLUX.1 VAE (ae) ae.safetensors | models/vae | 0.3 GB | required | Download → |
The files the official ComfyUI workflows load next to the model. Sizes read from Hugging Face (2026-10-01).
07Where the files go in ComfyUI
| File | Folder | Loader node |
|---|---|---|
| Diffusion model (.safetensors: 16-bit, FP8, INT8) | ComfyUI/models/diffusion_models | Load Diffusion Model |
| GGUF file (.gguf) | ComfyUI/models/unet | Unet Loader (GGUF) - from the ComfyUI-GGUF node pack |
| Text encoder | ComfyUI/models/text_encoders | Load CLIP / DualCLIPLoader (or the GGUF versions) |
| VAE | ComfyUI/models/vae | Load VAE |
08Measured and reported results
| Label | GPU | Setup | Result | Peak VRAM | Date | Source |
|---|---|---|---|---|---|---|
| measured | RTX 5060 Ti 16 GB | z_image_turbo_bf16.safetensors · 1024x1024 · 8 steps 3 timed runs after 1 warm-up (15.4 s incl. loading). Qwen3 4B FP8 text encoder + ae VAE, res_multistep/simple, CFG 1, shift 3. Time = sampling + VAE decode. ComfyUI 0.25.0 portable | 11 s / image | 14.6 GB | 2026-09-27 | |
| reported | RTX 3060 12 GB | 1024x1024 “私の環境（グラボ：RTX 3060 12GB）では解像度1024*1024pxの画像を生成するのに大体45秒程度かかりました” ComfyUI; approximate ('大体…程度'); file/steps not stated (article notes ~12GB model) | 45 s / image | — | 2025-12-05 | kurokumasoft.com → |
| reported | RTX 3090 24 GB | Tongyi-MAI/Z-Image-Turbo · 1024x1024 “NVIDIA RTX 3090 | 0.50it/s | 2.01s/it | CUDA 12.6 | ComfyUI (5151cff) | Arch Linux” Community speed list; ComfyUI, Euler/Normal, CFG 8, square 1:1, batch speed only (no per-image time); date = list last-updated date; Z-Image 1024px section; CFG 8 (above Turbo's us | 2.01 s/it | — | 2025-11-29 | huggingface.co → |
| reported | RTX 5060 Ti 16 GB | z_image_turbo_bf16.safetensors · 1328x1328 · 8 steps “約 35秒かかりました。一度モデルを VRAM にロードした後、プロンプトを変えての再実行だと約 22秒で生成できます。” ComfyUI; ~35 s first run incl. load, ~22 s warm; approximate ('約') | 22 s / image | — | 2025-11-30 | iwannacreateapps.com → |
| reported | RTX 5090 32 GB | fp8_e4m3fn (per log) “without SA: 0.95 seg” 'seg' = segundos (seconds) per image; SageAttention variants gave 0.87-0.91 s; resolution/steps not stated | 0.95 s / image | — | 2025-11-27 | github.com → |
Other people's numbers are copied as published, with a link. Send yours.
09Training a LoRA for Z-Image Turbo
| Trainer | VRAM | Type | Settings and quote | Source |
|---|---|---|---|---|
| SimpleTuner | 10 GB | stated minimum | Guide covers base + turbo (example trains Turbo): NF4/int8 base, 512px, batch 1, Lion8bit paged, gradient checkpointing; rank-16 NF4 ~10-12 GB, int8 ~16-24 GB, bf16 ~32-40 GB “the absolute minimum is a single 3080 10G (with aggressive quantisation/offload)” | github.com → |
reported From each trainer's own docs, 2026-09-25. Why the figures differ →
10Every file tracked for Z-Image Turbo
| File | Type | Size | Repo |
|---|---|---|---|
| z_image_turbo_bf16.safetensors | BF16 | 12.31 GB | Comfy-Org/z_image_turbo → |
| z_image_turbo_int8_convrot.safetensors | INT8 | 6.20 GB | Comfy-Org/z_image_turbo → |
| z_image_turbo_nvfp4.safetensors | NVFP4 | 4.51 GB | Comfy-Org/z_image_turbo → |
| z_image_turbo-Q8_0.gguf | Q8_0 | 7.22 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q6_K.gguf | Q6_K | 5.91 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q5_K_M.gguf | Q5_K_M | 5.52 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q5_K_S.gguf | Q5_K_S | 5.19 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q4_K_M.gguf | Q4_K_M | 4.98 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q4_K_S.gguf | Q4_K_S | 4.66 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q3_K_M.gguf | Q3_K_M | 4.12 GB | jayn7/Z-Image-Turbo-GGUF → |
| z_image_turbo-Q3_K_S.gguf | Q3_K_S | 3.79 GB | jayn7/Z-Image-Turbo-GGUF → |
| z-image-turbo-BF16.gguf | BF16 | 12.31 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-F16.gguf | F16 | 12.31 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q8_0.gguf | Q8_0 | 7.22 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q6_K.gguf | Q6_K | 5.91 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q5_K_M.gguf | Q5_K_M | 5.57 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q5_1.gguf | Q5_1 | 5.53 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q5_0.gguf | Q5_0 | 5.26 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q5_K_S.gguf | Q5_K_S | 5.24 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q4_K_M.gguf | Q4_K_M | 5.02 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q4_1.gguf | Q4_1 | 4.85 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q4_K_S.gguf | Q4_K_S | 4.71 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q4_0.gguf | Q4_0 | 4.59 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q3_K_M.gguf | Q3_K_M | 4.19 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q3_K_S.gguf | Q3_K_S | 3.95 GB | unsloth/Z-Image-Turbo-GGUF → |
| z-image-turbo-Q2_K.gguf | Q2_K | 3.64 GB | unsloth/Z-Image-Turbo-GGUF → |
Tongyi-MAI (Alibaba) 6B S3-DiT, 8-step distilled. Comfy-Org repo has ~5.2M downloads (hugely popular). No official FP8 in Comfy-Org repo (int8_convrot and nvfp4 instead). Text encoder Qwen3-4B (see qwen3-4b entry). A non-distilled base Z-Image was released 2026-01 (Comfy-Org/z_image) - not fetched.