by Stability AI
Two-stage cascade pipeline from Stability AI using Wurstchen architecture. Stage C (~3.6B) generates in a very small latent space, then Stage B (~1.5B) decodes to full resolution. More VRAM-efficient than single-stage models of similar quality.
VRAM requirements, GPU fit, and setup notes for Stable Cascade, including 8GB/12GB fit guidance where relevant. Recommended runtimes: ComfyUI and Diffusers support. Best download size: ~14.4 GB at FP16.
Your hardware
Detecting...
Measured quality metrics for Stable Cascade outputs.
How often humans prefer this model's output (0-100%)
Visual quality and composition rating (5-9 scale)
Compare which GPUs can run Stable Cascade at different precisions. FP8 uses less memory than FP16 when available, and the grade shows how comfortably each GPU handles the workload.
| Resolution | VRAM Required | RTX 4090 24GB | RTX 3060 12GB | RTX 4060 8GB | MacBook Pro M4 Pro 24GB |
|---|---|---|---|---|---|
| 512×512 | 15.3 GB | S● | D● | F● | A● |
| 768×768 | 15.6 GB | S● | D● | F● | A● |
| 1024×1024 | 16.0 GB | S● | D● | F● | B● |
Turbo / LCM distillation
Use distilled scheduler at 4-8 steps for faster iteration
ControlNets available
Add guided generation with 2 adapters (+1.5 GB VRAM each)
from diffusers import StableCascadeCombinedPipeline
import torch
pipe = StableCascadeCombinedPipeline.from_pretrained(
"stabilityai/stable-cascade",
torch_dtype=torch.float16
)
pipe.to("cuda")
image = pipe(
prompt="your prompt here",
num_inference_steps=20,
guidance_scale=4.0,
height=1024,
width=1024,
).images[0]
image.save("output.png")Get started
Setup instructions for running Stable Cascade locally
1. Download the model
Get the checkpoint from HuggingFace
2. Place in:
ComfyUI/models/checkpoints/3. Launch ComfyUI
python main.pyVRAM allocation at 1024×1024 on RTX 4090 24GB (24 GB)
Time per image at 1024×1024, 28 steps, FP16.
Download Stable Cascade in the precision that matches your GPU. Lower precision usually means less VRAM pressure, while higher precision keeps more quality.
| Format | Precision | Size | Provider | |
|---|---|---|---|---|
| safetensorsRecommended | FP16 | 14.4 GB | official | Download |
2 ControlNets available for Stable Cascade. ControlNets add guided image generation (edges, depth, pose) at the cost of extra VRAM.
Official Stability AI canny ControlNet for Stable Cascade Stage C.
Official Stability AI inpainting ControlNet for Stable Cascade Stage C.
Very few LoRAs available. Stable Cascade was largely eclipsed by SD 3.x and Flux.
Frequently asked questions
Stable Cascade (3.6B parameters) requires approximately 16.0 GB of VRAM at FP16 precision for standard 1024×1024 image generation. If you want a lighter setup, lower precisions like FP8 can reduce memory use when available.
Stable Cascade usually needs more than 8GB for comfortable local use. Check the VRAM table above for the exact resolution and precision trade-off.
Stable Cascade is marked for ComfyUI and Diffusers support in our catalog, so those are the runtimes we recommend first for local setup. If your workflow uses another front end, check the model's available formats and workflow notes above before downloading.
Yes, the RTX 4090 (24 GB VRAM) can run Stable Cascade comfortably at FP16. Expected generation time is around ~10s per image at 1024×1024.
Yes, Stable Cascade has 2 ControlNet adapters available: Canny Edge, Inpainting. Each ControlNet adds roughly 1.5 GB of extra VRAM.
Very few LoRAs available. Stable Cascade was largely eclipsed by SD 3.x and Flux. The LoRA ecosystem for Stable Cascade is rated as "minimal". Each LoRA adds roughly 0.2 GB of extra VRAM.
On a reference GPU (RTX 4090 24GB), Stable Cascade generates a 1024×1024 image in approximately ~10s at FP16 with 28 inference steps. Faster GPUs with higher memory bandwidth will produce images more quickly.
See also