Stable Diffusion

An open-source text-to-image generation model that runs locally and can be fine-tuned for commercial use — the cornerstone of the open-source text-to-image ecosystem.

Stable Diffusion is the open-source text-to-image generation model from Stability AI and one of the most important open-source projects of the entire AI image generation wave. Users enter a text prompt, and the model generates a matching image; with extension technologies such as LoRA, ControlNet, and IP-Adapter, it can also achieve style transfer, pose control, character consistency, and inpainting. Compared with closed-source services like Midjourney, Stable Diffusion’s biggest advantage is that it “runs locally, can be fine-tuned, and can be used commercially” — which has given rise to a massive community ecosystem including Civitai, ComfyUI, and Stable Diffusion WebUI.

Pricing follows a “free open source + commercial services” model. The model weights themselves are free to download and run locally; Stability AI officially offers the DreamStudio web tool and API access, billed by the number of generations; enterprise customers can also obtain compliance support through Amazon Bedrock and Stability AI’s enterprise licensing. For creators and small teams, local deployment plus community models is essentially zero-cost; for commercial customers that need SLAs and copyright assurance, the official API or Bedrock is the better fit.

As for the company’s current state, Stability AI did indeed go through a leadership change and strategic restructuring around 2024, but since 2025, under new CEO Prem Akkaraju, it has refocused on enterprise-grade creative services. Between 2025 and 2026, Stability AI successively formed strategic partnerships with Electronic Arts, Warner Music Group, and Universal Music Group, brought its image services onto Amazon Bedrock, and released Stable Audio 3.0 and the new Stable Diffusion 3.5 model series. As of mid-2026, Stability AI remains fully operational and continues to release models.

In terms of its niche, Stable Diffusion is the leading representative of “controllable, customizable, and self-hostable” image generation. Midjourney is better suited for everyday users pursuing artistic results, DALL·E for integration with the OpenAI ecosystem, while Stable Diffusion is the top choice for designers, developers, advertising agencies, and game studios that need deep customization.

Note: Running Stable Diffusion locally requires a certain amount of VRAM (8GB+ recommended), and ComfyUI is recommended for complex workflows. For commercial use, pay attention to Stability AI’s license terms — some derivative models may carry additional restrictions.

Model Iteration Timeline

Time	Version/Model	Key Points
2022-08	Stable Diffusion 1.x	First open-source text-to-image model to go viral
2022-11	SD 1.5	Community ecosystem took off; LoRA/ControlNet emerged
2023-07	SDXL	Native 1024px resolution, a leap in image quality
2024-06	Stable Diffusion 3	New architecture with enhanced text rendering
2024-10	Stable Diffusion 3.5	Multiple model sizes covering both enterprise and consumer needs
2025-05	SD 3.5 AMD NPU Edition	Optimized for local inference on AMD NPUs

Source: Agentopia · Site-wide machine-readable catalog at /llms.txt