Can the RTX 5090 32 GB run Z-Image Turbo?
Yes. 16-bit fits on the RTX 5090 32 GB with 17.7 GB to spare.
16-bit (12.3 GB) needs about 14.3 GB in use, so 17.7 GB of the 32 GB stays free: room for a larger image, a LoRA or a ControlNet.
01What to download
| Part | File | Folder | Size | |
|---|---|---|---|---|
| Model | z_image_turbo_bf16.safetensors 16-bit · Comfy-Org/z_image_turbo | models/diffusion_models | 12.3 GB | Download → |
| Text encoder | qwen_3_4b_fp8_mixed.safetensors Qwen3 4B FP8 · Comfy-Org/z_image_turbo | models/text_encoders | 5.6 GB | Download → |
| VAE | ae.safetensors FLUX.1 VAE (ae) · Comfy-Org/z_image_turbo | models/vae | 0.3 GB | Download → |
| Total download · keep about the same free on disk | 18.3 GB | |||
Alternatives: Qwen3 4B BF16 (8.0 GB). Sizes from Hugging Face, 2026-10-01.
Free ComfyUI workflow, tested on my RTX 5060 Ti: 16-bit workflow (.json). Drag the file into ComfyUI. All workflows →
02Every Z-Image Turbo file on 32 GB
| File | Size | Needed | On this card | Quality | Download |
|---|---|---|---|---|---|
| 16-bit ← SAFETENSORS · Comfy-Org | 12.3 GB | 14.3 GB | fits · 17.7 GB spare | the original weights | Hugging Face → |
| INT8 SAFETENSORS · Comfy-Org | 6.2 GB | 8.2 GB | fits · 23.8 GB spare | practically identical to the original | Hugging Face → |
| Q8_0 GGUF · unsloth | 7.2 GB | 9.2 GB | fits · 22.8 GB spare | practically identical to the original | Hugging Face → |
| Q6_K GGUF · unsloth | 5.9 GB | 7.9 GB | fits · 24.1 GB spare | very close to the original | Hugging Face → |
| Q5_K_M GGUF · unsloth | 5.6 GB | 7.6 GB | fits · 24.4 GB spare | close; small differences in fine detail | Hugging Face → |
| Q4_K_M GGUF · unsloth | 5.0 GB | 7.0 GB | fits · 25.0 GB spare | good; some loss in fine detail and text | Hugging Face → |
| Q3_K_M GGUF · unsloth | 4.2 GB | 6.2 GB | fits · 25.8 GB spare | noticeable loss of detail | Hugging Face → |
| Q2_K GGUF · unsloth | 3.6 GB | 5.6 GB | fits · 26.4 GB spare | heavy loss; a last resort | Hugging Face → |
“Needed” adds 1.2 GB of working memory for Z-Image Turbo and a 0.8 GB reserve. Sizes 2026-10-01.
03Measured and reported results
| Label | Setup | Result | Peak VRAM | Date | Source |
|---|---|---|---|---|---|
| reported | 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.
04About the RTX 5090 32 GB
The RTX 5090 32 GB is a Blackwell GPU with FP8 and FP4 hardware: ComfyUI computes Comfy-Org's FP8 files natively here, and NVFP4 files (where a model offers them) are faster still.
32 GB GDDR7 · 512-bit · 1792 GB/s · Blackwell. Everything that runs on it →
05Questions
How much VRAM does Z-Image Turbo need?
Around 5.6 GB with the smallest file (Q2_K) and 8.2 GB with an 8-bit file (INT8). The full 16-bit file needs about 14.3 GB.
Which Z-Image Turbo file should I download for the RTX 5090 32 GB?
16-bit (12.3 GB) from Comfy-Org/z_image_turbo.
How much do I need to download for Z-Image Turbo on the RTX 5090 32 GB?
18.3 GB: z_image_turbo_bf16.safetensors, qwen_3_4b_fp8_mixed.safetensors, ae.safetensors.