image model · stable-diffusion · Linux
Can I run Stable Diffusion 3.5 Large on Nvidia DGX Spark (128GB)?
Yes. Stable Diffusion 3.5 Large runs on Nvidia DGX Spark (128GB) at Q4 GGUF (~7 GB of ~112 GB usable).
Runs at Q4 GGUF using ~7 GB of ~112 GB usable.
- Peak VRAM
- ~7 GB
- Usable on device
- ~112 GB
- Device memory
- 128 GB
- Quant
- Q4 GGUF
How to run it
Use ComfyUI or Draw Things at Q4 GGUF. The big text encoder is loaded to encode your prompt, then offloaded before generation, which is why peak VRAM stays near the backbone size rather than the sum of every file.
- Type
- image (MMDIT)
- Parameters
- 8.1B
- Peak VRAM
- ~7 GB at Q4 GGUF
- Resolution
- 1024×1024
- License
- Stability Community License
- Memory
- 128 GB unified
- Usable for weights
- ~112 GB
- Power draw
- ~140 W
- Best runtime
- llama.cpp (CUDA) / Ollama (CUDA)
You could also run
Run Stable Diffusion 3.5 Large on other hardware
FAQ
Can Nvidia DGX Spark (128GB) run Stable Diffusion 3.5 Large?
Yes. Stable Diffusion 3.5 Large runs on Nvidia DGX Spark (128GB) at Q4 GGUF (~7 GB of ~112 GB usable).
How much VRAM does Stable Diffusion 3.5 Large need?
Nvidia DGX Spark (128GB) has room to spare. At Q4 GGUF the realistic peak is ~7 GB of VRAM, versus ~19 GB with every component kept resident (no offload). With aggressive CPU offload it drops to ~5 GB, much slower.
What do I use to run Stable Diffusion 3.5 Large locally?
Stable Diffusion 3.5 Large runs in ComfyUI or Draw Things (among others). It loads as a diffusion checkpoint plus its text encoder and VAE, not a single chat command.
Sources
-
docs.nvidia.com · 2 sources
-
github.com · 1 source
-
huggingface.co · 2 sources
-
nvidia.com · 1 source
-
nvidianews.nvidia.com · 1 source
-
ollama.com · 1 source
-
tomshardware.com · 1 source
VRAM figures are sourced peak-usage anchors at the noted quant; catalog updated 2026-11-01. See methodology.