Stable Diffusion 3.5 Large

Generates images from natural language text prompts using a Multimodal Diffusion Transformer (MMDiT) architecture with 8.1 billion parameters. The model operates in latent space, progressively denoising from random noise conditioned on text embeddings across transformer blocks with integrated Query-Key Normalization. Supports output resolutions from 512×512 to 1 megapixel, with claimed superior text rendering and prompt adherence compared to Stable Diffusion 3.0.

Stable Diffusion 3.5 Large Turbo variant generates images in 4 diffusion steps instead of the standard multi-step process, achieving 'considerably faster' inference while maintaining the 8.1B parameter architecture. Uses knowledge distillation techniques to compress the denoising schedule without retraining from scratch, trading marginal quality for speed. Designed for real-time or interactive applications where latency is critical.

Stability AI provides inference code on GitHub (repository URL not specified in documentation) enabling self-hosted deployment on various hardware configurations and frameworks. Code supports PyTorch and likely other inference engines (e.g., ONNX, TensorRT). No proprietary inference runtime required; standard Python/PyTorch stack enables deployment on cloud VMs, on-premises servers, or edge devices. Inference code is open-source, enabling community optimization and integration.

Achieves improved text rendering quality compared to predecessor models (SD 3 Medium) through the MMDiT architecture's joint text-image processing and enhanced text embedding integration. The model can generate readable, correctly-spelled text within images at various sizes and styles, addressing a major limitation of prior diffusion models that struggled with text generation.

Demonstrates enhanced ability to follow detailed prompts and understand complex compositional requirements through the MMDiT architecture's improved text-image alignment and larger effective context window. The model better interprets spatial relationships, object interactions, and nuanced prompt specifications compared to prior diffusion models, reducing need for prompt engineering and negative prompts.

Stable Diffusion 3.5 Medium variant reduces model size to 2.5 billion parameters while maintaining MMDiT architecture, enabling inference 'out of the box' on consumer hardware without GPU optimization. Uses improved MMDiT-X architecture design to maximize parameter efficiency. Supports output resolutions from 0.25 to 2 megapixels, doubling the maximum resolution of the Large variant while reducing memory footprint.

Supports Low-Rank Adaptation (LoRA) fine-tuning on all model variants (Large, Large Turbo, Medium) with stabilized training process via Query-Key Normalization in transformer blocks. LoRA adds learnable low-rank matrices to attention weights without modifying base model weights, enabling efficient adaptation to custom styles, objects, or domains. Designed as primary customization mechanism with documented support for community-contributed LoRA modules.

Model weights released under Stability AI Community License as open-source artifacts, available for download from Hugging Face in standard formats (likely safetensors or PyTorch). License explicitly permits commercial and non-commercial use, fine-tuning, redistribution, and monetization of derived works across the entire pipeline (fine-tuned models, LoRA modules, applications, artwork). No API key or proprietary access required; full model control and deployment flexibility.

Stability AI Brand Studio provides a SaaS platform offering web UI and workflow tools for image generation, inpainting, outpainting, and background removal. Implements 'Curated Model Routing' that selects from multiple providers (including Stable Diffusion variants) based on task requirements. Tiered pricing model: free trial (1000 credits), Core ($50/month, 5000 credits/month), and Enterprise (custom). Abstracts model selection and infrastructure management from users.

Stable Diffusion 3.5 Large supports output resolutions from 512×512 to 1 megapixel (1,000,000 pixels), enabling generation of images suitable for print, large displays, or detailed crops. Latent diffusion architecture operates in compressed latent space, enabling efficient generation of high-resolution outputs without proportional VRAM increase. Supports arbitrary aspect ratios within resolution constraints (e.g., 1024×1024, 768×1280, 512×1920).

Stable Diffusion 3.5 Large claims 'superior text rendering' compared to predecessors through improved MMDiT architecture and training. Text-to-image conditioning operates across all transformer blocks with Query-Key Normalization, enabling tighter coupling between text tokens and image generation. Supports rendering of multi-word phrases, proper spelling, and text layout within images, addressing a known weakness of earlier diffusion models.

Stable Diffusion 3.5 Large claims 'exceptional prompt adherence' and 'improved compositional understanding' through MMDiT architecture that jointly processes text and image tokens. Transformer blocks with Query-Key Normalization enable better spatial reasoning about object relationships, counts, and layout. Supports complex prompts describing multiple objects, their spatial relationships, and attributes without degradation in quality.

Supports integer seed parameter to control randomness in image generation, enabling reproducible outputs and intentional variation. Same prompt with same seed produces identical image; different seeds produce diverse outputs from the same prompt. Model intentionally preserves variation across seeds to maintain knowledge base diversity and prevent mode collapse, documented as design trade-off.