SD 3.5 Large — implementation by Spacelike AI

Multimodal Diffusion Transformer · 8.1 B parameters · Stability AI

SD 3.5 Large final-sample output generated by Spacelike AI.

SD 3.5 Large is the flagship of the SD 3.5 family. At 8.1 B parameters it is the largest MM-DiT-based Stable Diffusion model Stability AI has released, targeting professional image generation at 1-megapixel native resolution with substantially improved typography, anatomy and compositional coherence over SD 3.

A sibling distilled variant, SD 3.5 Large Turbo, produces equivalent quality in 4 steps at the cost of some prompt adherence. Both share the same MM-DiT backbone.

Specification

Architecture
MM-DiT · 38 blocks · improved-QK-norm
Parameters
8.1 B (backbone)
Training objective
Rectified flow matching
Native resolution
1024 × 1024 (1 MP)
Text encoders
CLIP-L + CLIP-G + T5-XXL (frozen)
Sampler shown
Euler · 28 steps · cfg 3.5
License
Stability AI Community License
Release
October 22, 2024
Checkpoint
sd3.5_large.safetensors

Client: Tenstorrent Inc. — performance improvement.

implementation · Hugging Face · Vendor announcement

Live sample on the Spacelike AI home page.

Sample images on this page are licensed under CC BY 4.0 — reuse with attribution to Spacelike AI and a link back to spacelike.ai.

SpacelikeAI Pushing AI Models To The Limits Of Hardware
01
Denoising Step / 05
sigma 14.6 · latent noise
cfg7.5
step1 / 5
seed0x7A3F
Loading models…