Datasheet for local AI49 models98 GPUsData read 2026-09-25
No adsNo tracking
Check my GPU →

Stable Diffusion 3.5 Large on 12 GB of VRAM

Image model · 8.1BAny 12 GB GPUData 2026-09-25
Runs well
Yes — and comfortably.

Download Q8_0 (8.8 GB). With the model's working memory it needs about 11.1 GB, leaving 0.9 GB spare on 12 GB. Quality: practically identical to the original.

Best fileQ8_0
File size8.8 GB
VRAM needed~11.1 GB
System RAM32 GB+
Memory map · 12 GB GPU11.1 GB / 12 GB
06 GB12 GB
Weights Q8_0 · 8.8 GBWorking memory · 1.5 GBReserve · 0.8 GBFree · 0.9 GB
calculated from real file sizes plus working memory. Not yet measured on this setup. How this works.

01What to download

PartFileFolderSize
Modelsd3.5_large-Q8_0.gguf
Q8_0 · city96/stable-diffusion-3.5-large-gguf
models/unet8.8 GBDownload →
Text encodert5xxl_fp8_e4m3fn.safetensors
T5-XXL FP8 · Comfy-Org/stable-diffusion-3.5-fp8
models/text_encoders4.9 GBDownload →
Text encoderclip_l.safetensors
CLIP-L · Comfy-Org/stable-diffusion-3.5-fp8
models/text_encoders0.2 GBDownload →
Text encoderclip_g.safetensors
CLIP-G · Comfy-Org/stable-diffusion-3.5-fp8
models/text_encoders1.4 GBDownload →
Total download · keep about the same free on disk15.3 GB

Alternatives: T5-XXL FP16 (9.8 GB). The 16-bit text encoder is a little more faithful; the smaller one is chosen here because it loads faster and needs less RAM. Sizes read from Hugging Face (2026-09-25). “Download” links start the file directly; the file name links are the same files the official ComfyUI workflows use.

System RAM: 32 GB or more. ComfyUI keeps the model file, the text encoder and the VAE in system RAM and moves them to the GPU as needed. With this set of files that is about 15.3 GB, plus roughly 6 GB for Windows, ComfyUI and a browser: 21.3 GB in total. With less RAM it still runs, but Windows starts swapping to disk and loading gets very slow. calculated

02Every file, on 12 GB

FileSizeNeededOn this cardQualityDownload
16-bit
SAFETENSORS · stabilityai
16.5 GB18.8 GB6.8 GB too bigthe original weightsHugging Face →
Q8_0 ←
GGUF · city96
8.8 GB11.1 GBfits · 0.9 GB sparepractically identical to the originalHugging Face →
Q5_1
GGUF · city96
6.3 GB8.6 GBfits · 3.4 GB spareclose; small differences in fine detailHugging Face →
Q4_1
GGUF · city96
5.3 GB7.6 GBfits · 4.4 GB sparegood; some loss in fine detail and textHugging Face →

Sizes from the Hugging Face file listing, read 2026-09-25. “Needed” = file + 1.5 GB working memory for this model + 0.8 GB kept free for the system.

03The text encoder

CLIP-L + CLIP-G + T5-XXL: 9.8 GB as 16-bit, 4.9 GB as FP8 (T5-XXL sizes shown; CLIP-L and CLIP-G add about 1.6 GB). ComfyUI encodes the prompt first and can push the encoder out of VRAM before sampling, so it does not have to fit together with the model. On 12 GB the FP8 encoder fits on its own, so prompt encoding stays fast.