How much VRAM does Stable Diffusion 3.5 Large need?

Stable Diffusion 3.5 Large requires 12.5 GB VRAM at recommended quality (FP16). At lower quality settings, it can fit in as little as 9 GB.

What is the best GPU for Stable Diffusion 3.5 Large?

The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Stable Diffusion 3.5 Large, achieving 12 tok/s at FP16 with an excellent rating.

Can I run Stable Diffusion 3.5 Large on an RTX 4060 Ti?

Yes. On the NVIDIA GeForce RTX 4060 Ti 16GB, Stable Diffusion 3.5 Large runs at – (FP16, good).

What quantization should I use for Stable Diffusion 3.5 Large?

For the best quality, use FP16 (12.5 GB VRAM). If your GPU has limited VRAM, Q8_0 (9 GB) is the most efficient option with acceptable quality.

Image gen8.1B

Image gen

Stable Diffusion 3.5 Large

Stable Diffusion · Stability AI Community License

Latest Stable Diffusion architecture with significantly improved prompt adherence, typography, and composition over SDXL. Higher VRAM requirement (12GB+ at FP16) but produces noticeably better images. The current standard for local image generation.

Parameters: 8.1B
Architecture: Dense
Context: 77 tokens
Released: 2024-10-22
Engines: ComfyUI, AUTOMATIC1111, InvokeAI, diffusers

Parameters

8.1B

VRAM

12.5 GB

Context

Formats

GPUs

Stable Diffusion 3.5 Large (8.1B) requires 12.5 GB VRAM at recommended quality (FP16). At efficient quality (Q8_0), it fits in 9 GB VRAM, making it compatible with the NVIDIA RTX 4080 Laptop (120-150W). On NVIDIA Grace Blackwell Ultra GB300, expect approximately 12 tok/s at FP16. For the best experience, Budget Home AI Server ($1,162) is recommended.

Source: OwnRig methodology

VRAM (Recommended)

12.5 GB

Quantization

FP16

File Size

11.8 GB

Max Context

0K tokens

Primary Use

Image gen

Memory

VRAM Requirements

Quality	Quantization	VRAM	File Size
full	FP16	12.5 GB	11.8 GB
efficient	Q8_0	9 GB	8.5 GB

Compatible GPUs

43 devices


NVIDIA Grace Blackwell Ultra GB300	FP16	12 tok/s	Excellent
NVIDIA GeForce RTX 3090	FP16	–	Excellent
NVIDIA GeForce RTX 4070 Ti Super	FP16	–	Excellent
NVIDIA GeForce RTX 4080 Super	FP16	–	Excellent
NVIDIA GeForce RTX 4090	FP16	–	Excellent
NVIDIA GeForce RTX 5080	FP16	–	Excellent
NVIDIA GeForce RTX 5090	FP16	–	Excellent
NVIDIA RTX PRO 6000 Blackwell	FP16	–	Excellent
NVIDIA RTX PRO 6000 Blackwell Max-Q	FP16	–	Excellent
Apple M4 Max (128GB Unified)	FP16	–	Good
Apple M4 Max (36GB Unified)	FP16	–	Good
Apple M4 Max (64GB Unified)	FP16	–	Good
Apple M4 Ultra (192GB)	FP16	–	Good
NVIDIA GeForce RTX 3060 12GB	Q8_0	–	Good
NVIDIA GeForce RTX 4060 Ti 16GB	FP16	–	Good
NVIDIA GeForce RTX 4070 Super	Q8_0	–	Good
NVIDIA GeForce RTX 4070 Ti 12GB	Q8_0	–	Good
NVIDIA RTX 4080 Laptop (120-150W)	Q8_0	–	Good
NVIDIA RTX 4090 Laptop (150-175W)	FP16	–	Good
AMD Radeon RX 7900 XTX	FP16	–	Good
NVIDIA GeForce RTX 5060 Ti 16GB	FP16	–	Good
Apple M4 (16GB Unified)	FP16	–	Acceptable
Apple M4 Pro (24GB Unified)	FP16	–	Acceptable
Apple M4 Pro (48GB)	FP16	–	Acceptable
NVIDIA GeForce RTX 3080 10GB	Q8_0	–	Acceptable
AMD Radeon Pro W7900	FP16	–	Acceptable
AMD Radeon RX 9070	FP16	–	Acceptable
Apple M1 (16GB Unified)	FP16	–	Acceptable
Apple M1 Pro (16GB Unified)	FP16	–	Acceptable
Apple M2 (16GB Unified)	FP16	–	Acceptable
Apple M2 Pro (16GB Unified)	FP16	–	Acceptable
Apple M3 (16GB Unified)	FP16	–	Acceptable
AMD Radeon RX 9060 XT 16GB	FP16	–	Acceptable
Apple M3 Pro (18GB Unified)	FP16	–	Marginal
AMD Radeon RX 7600	Q8_0	2 tok/s	Marginal
NVIDIA GeForce RTX 4060 8GB	Q8_0	–	Not viable
NVIDIA RTX 4060 Laptop (40-60W)	Q8_0	–	Not viable
NVIDIA RTX 4070 Laptop (80-115W)	Q8_0	–	Not viable
Apple M1 (8GB Unified)	FP16	–	Not viable
Apple M2 (8GB Unified)	FP16	–	Not viable
Apple M3 (8GB Unified)	FP16	–	Not viable
AMD Radeon RX 9060 XT 8GB	FP16	–	Not viable
NVIDIA GeForce RTX 5060 8GB	Q8_0	–	Not viable

Showing 43 of 43 entries

Hardware

Recommended Builds

Complete PC builds that can run Stable Diffusion 3.5 Large.

High-end

AMD AI Powerhouse

24 GB of AI power at nearly half the NVIDIA price

RX 7900 XTX 24GB·24 GBVRAM

Runs 7 models

$1,699

FAQ

Frequently Asked Questions

How much VRAM does Stable Diffusion 3.5 Large need?: Stable Diffusion 3.5 Large requires 12.5 GB VRAM at recommended quality (FP16). At lower quality settings, it can fit in as little as 9 GB.
What is the best GPU for Stable Diffusion 3.5 Large?: The NVIDIA Grace Blackwell Ultra GB300 delivers the best performance for Stable Diffusion 3.5 Large, achieving 12 tok/s at FP16 with an excellent rating.
Can I run Stable Diffusion 3.5 Large on an RTX 4060 Ti?: Yes. On the NVIDIA GeForce RTX 4060 Ti 16GB, Stable Diffusion 3.5 Large runs at – (FP16, good).
What quantization should I use for Stable Diffusion 3.5 Large?: For the best quality, use FP16 (12.5 GB VRAM). If your GPU has limited VRAM, Q8_0 (9 GB) is the most efficient option with acceptable quality.

Related Guides

Buying Guide

Best GPUs for Stable Diffusion, Flux, and SD3 in 2026

GPU requirements for SDXL, Stable Diffusion 3 Medium, SD 3.5 Large, and FLUX.1 Dev. Per-GPU performance verdicts for RTX 4060 Ti, RTX 4070, RTX 4090, and Apple Silicon.

All models

Data confidence: estimated. Source

VRAM requirements are calculated from model parameters and may vary by inference engine, context length, and batch size. Performance estimates are based on community benchmarks and should be verified for your specific configuration.Stable Diffusion is a trademark of its respective owner. OwnRig is not affiliated with or endorsed by the model creator.