Stable Diffusion 3.5 Large VRAM Requirements
Stable Diffusion 3.5 Large needs about 8.6 GB of VRAM with FP8 weights at 1024×1024, loading one model at a time. In FP16 it needs 16.2 GB (29.6 GB with everything loaded), and GGUF Q4 brings it to 6.4 GB. The smallest consumer GPU that runs it is the RTX 3050 8GB with Q4 weights, or the Intel Arc B570 10GB in FP8.
8.6 GB peak VRAM
Stable Diffusion 3.5 Large at 1024×1024 with FP8 weights, loading one model at a time (ComfyUI default on smaller cards). Keeping everything loaded needs 16.8 GB.
- Diffusion model 7.5 GB
- Text encoders 5.2 GB
- VAE 0.2 GB
- Peak activations 3.1 GB
- Overhead 0.8 GB
GPU fit
- RTX 4060 8GB87% on GPU
- RTX 3060 12GBFitswith swapping
- RTX 4060 Ti 16GBFitswith swapping
- RTX 4090 24GBFitsall loaded
- RTX 5090 32GBFitsall loaded
- RTX 6000 Ada 48GBFitsall loaded
- RTX PRO 6000 Blackwell 96GBFitsall loaded
- Apple M4 Max 128GBFitsall loaded
The VAE decode is the peak here. Turn on tiled VAE decoding to cut it to about 0.8 GB.
Stable Diffusion 3.5 Large VRAM by precision
At 1024×1024. Text encoders in FP8 except for the FP16 row.
| Weights | Diffusion model | Peak (swapping) | All loaded | Smallest tier |
|---|---|---|---|---|
| FP16 / BF16 | 15.1 GB | 16.2 GB | 29.6 GB | 20 GB |
| FP8 | 7.5 GB | 8.6 GB | 16.8 GB | 10 GB |
| GGUF Q8_0 | 8.0 GB | 9.1 GB | 17.3 GB | 10 GB |
| GGUF Q5_K_S | 5.2 GB | 6.4 GB | 14.5 GB | 8 GB |
| GGUF Q4_K_S / NF4 | 4.3 GB | 6.4 GB | 13.6 GB | 8 GB |
VRAM by resolution
FP8 weights, with model swapping.
| Resolution | Latent tokens | Activations | Peak VRAM |
|---|---|---|---|
| 1024×1024 | 4,096 | 0.3 GB | 8.6 GB |
| 1536×1536 | 9,216 | 0.7 GB | 9.0 GB |
About Stable Diffusion 3.5 Large
An 8.1B diffusion transformer. In FP16 the diffusion model and T5 together need more than 24 GB if kept loaded; FP8 weights and model swapping bring it to 12 to 16 GB cards.
| Component | Size |
|---|---|
| Diffusion model | 8.1B parameters (MMDiT) |
| Text encoders | CLIP ViT-L (0.123B), OpenCLIP ViT-bigG (0.695B), T5-XXL encoder (4.76B) |
| VAE | 0.084B parameters |
| Classifier-free guidance | Yes (2 passes per step) |
| Released | 2024-10 |
| License | Stability AI Community License |
Stable Diffusion 3.5 Large VRAM: common questions
What GPU do I need for SD 3.5 Large?
About 9 GB with FP8 weights and model swapping, so 12 GB cards work. FP16 with swapping needs about 16 GB, which means a 24 GB card in practice.
How much VRAM does Stable Diffusion 3.5 Large need?
At its default 1024×1024: about 16.2 GB in FP16 with model swapping (29.6 GB with everything loaded), 8.6 GB in FP8 and 6.4 GB with GGUF Q4 weights.