Stable Diffusion 3.5 Large
Stableby Stability AI
2.5B MMDiT transformer with triple text encoder (5.5B combined: T5-XXL 4.7B + CLIP-L 0.123B + OpenCLIP-G 0.695B). Improved text rendering and composition over SDXL.
VRAM requirements, GPU fit, and setup notes for Stable Diffusion 3.5 Large, including 8GB/12GB fit guidance where relevant. Recommended runtimes: ComfyUI and Diffusers support. Best download size: ~16.5 GB at FP16.
- MMDiT architecture — improved over UNet
- Triple text encoder (5.5B) for better prompt following
- Better text rendering than SDXL
- 2.5B MMDiT transformer — efficient architecture
Your hardware
Detecting...
Image Quality Benchmarks
Measured quality metrics for Stable Diffusion 3.5 Large outputs.
How often humans prefer this model's output (0-100%)
Visual quality and composition rating (5-9 scale)
Text-image alignment accuracy (higher is better)
VRAM Requirements by Resolution and Precision
Compare which GPUs can run Stable Diffusion 3.5 Large at different precisions. FP8 uses less memory than FP16 when available, and the grade shows how comfortably each GPU handles the workload.
FP16 (full precision)
| Resolution | VRAM Required | RTX 4090 24GB | RTX 3060 12GB | RTX 4060 8GB | MacBook Pro M4 Pro 24GB |
|---|---|---|---|---|---|
| 512×512 | 17.5 GB | S● | F● | F● | B● |
| 768×768 | 17.7 GB | S● | F● | F● | B● |
| 1024×1024 | 18.0 GB | S● | F● | F● | B● |
Optimization Tips
Turbo / LCM distillation
Use distilled scheduler at 4-8 steps for faster iteration
ControlNets available
Add guided generation with 1 adapter (+2.0 GB VRAM each)
Run with Python
from diffusers import StableDiffusion3Pipeline
import torch
pipe = StableDiffusion3Pipeline.from_pretrained(
"stabilityai/stable-diffusion-3.5-large",
torch_dtype=torch.float16
)
pipe.to("cuda")
image = pipe(
prompt="your prompt here",
num_inference_steps=28,
guidance_scale=7.0,
height=1024,
width=1024,
).images[0]
image.save("output.png")Get started
Setup instructions for running Stable Diffusion 3.5 Large locally
1. Download the model
Get the checkpoint from HuggingFace
2. Place in:
ComfyUI/models/checkpoints/3. Launch ComfyUI
python main.pyComfyUI Workflow
Basic txt2img workflow for Stable Diffusion 3.5 Large
Drag & drop into ComfyUI or use File → Import
Memory Breakdown
VRAM allocation at 1024×1024 on RTX 4090 24GB (24 GB)
Estimated Generation Time
Time per image at 1024×1024, 28 steps, FP16.
Sample Outputs
Available Formats, Downloads & Setup
Download Stable Diffusion 3.5 Large in the precision that matches your GPU. Lower precision usually means less VRAM pressure, while higher precision keeps more quality.
| Formato | Precisión | Tamaño | Proveedor | |
|---|---|---|---|---|
| safetensorsRecomendado | FP16 | 16.5 GB | official | Descargar |
ControlNet Support
1 ControlNet available for Stable Diffusion 3.5 Large. ControlNets add guided image generation (edges, depth, pose) at the cost of extra VRAM.
LoRA Ecosystem
LimitedVery few LoRAs available. SD 3.5 ecosystem is still in early stages.
Approximately 50 LoRAs available on CivitAI. Each LoRA adds ~0.2 GB VRAM.
Related Workflows
You might also like
Frequently asked questions
FAQ — Stable Diffusion 3.5 Large VRAM, Runtimes & Fit
How much VRAM does Stable Diffusion 3.5 Large need?
Stable Diffusion 3.5 Large (2.5B parameters) requires approximately 18.0 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.
Can I run Stable Diffusion 3.5 Large on an 8GB GPU?
Stable Diffusion 3.5 Large usually needs more than 8GB for comfortable local use. Check the VRAM table above for the exact resolution and precision trade-off.
Does Stable Diffusion 3.5 Large work in ComfyUI and Diffusers?
Stable Diffusion 3.5 Large 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.
Can I run Stable Diffusion 3.5 Large on RTX 4090?
Yes, the RTX 4090 (24 GB VRAM) can run Stable Diffusion 3.5 Large comfortably at FP16. Expected generation time is around ~22s per image at 1024×1024.
Does Stable Diffusion 3.5 Large support ControlNet?
Yes, Stable Diffusion 3.5 Large has 1 ControlNet adapter available: Canny Edge. Each ControlNet adds roughly 2.0 GB of extra VRAM.
Does Stable Diffusion 3.5 Large have LoRA support?
Very few LoRAs available. SD 3.5 ecosystem is still in early stages. The LoRA ecosystem for Stable Diffusion 3.5 Large is rated as "minimal". There are approximately 50 LoRAs available on Civitai. Each LoRA adds roughly 0.2 GB of extra VRAM.
How fast is Stable Diffusion 3.5 Large?
On a reference GPU (RTX 4090 24GB), Stable Diffusion 3.5 Large generates a 1024×1024 image in approximately ~22s at FP16 with 28 inference steps. Faster GPUs with higher memory bandwidth will produce images more quickly.
About Stable Diffusion 3.5 Large
See also