Model Family
Stability
Stability AI develops the Stable Diffusion family and related generative media models. Its image models are associated with open-weight, self-hostable, customizable, and API-accessible workflows, which makes Stability an important contrast to providers available mainly through hosted product and API surfaces.
3 models · 348 keywords · 3087 images on blupromptsModel Snapshot
Stable Diffusion 3.5 Large is the high-capability SD3.5 variant. If it becomes public on bluprompts, it should be framed as the stronger Stable Diffusion comparison for quality and prompt adherence, separate from Medium's more accessible workflow positioning.
- Best For
- stronger prompt adherence, higher-detail compositions, and professional-resolution outputs
- Pricing Note
- Pricing depends on platform and deployment route; self-hosting, Stability's API, and third-party platforms can each carry different cost assumptions.
- Release Context
- Stability describes SD3.5 Large as the most powerful model in the Stable Diffusion family at release, with 8.1B parameters and professional 1-megapixel use cases.
*As of June 27, 2026: SD3.5 Large follows the same Stability route-dependent pricing/licensing framing as SD3.5 Medium: self-hosted weights, Stability API, or third-party hosting, with Community License terms for qualifying non-commercial and commercial use.
Source: Stability AI SD3.5 announcement
Model Snapshot
Stable Diffusion 3.5 Medium represents the accessible side of the SD3.5 family in the published examples. It is smaller than Large, practical for consumer-hardware and customization-oriented workflows, and useful for comparing how Stable Diffusion interprets the same keyword scenes as hosted product and API providers.
- Best For
- efficient generation, customization-friendly workflows, consumer-hardware use, broad style exploration, and open/self-hostable comparison against hosted product and API providers
- Pricing Note
- Stability pricing depends on the generation route. Stable Diffusion 3.5 models can be used through self-hosted weights, Stability's API, or third-party hosting platforms, so public copy should not reduce Stability pricing to one per-image API number.
- Release Context
- Stable Diffusion 3.5 Medium is part of the SD3.5 family alongside Large and Large Turbo. Stability describes Medium as a 2.5B-parameter model designed to run out of the box on consumer hardware, balancing quality, customization, prompt adherence, and multi-resolution generation from 0.25 to 2 megapixels.
*As of June 27, 2026: Stability says SD3.5 model weights are available through Hugging Face, free for non-commercial use, and free for commercial use up to $1M in annual revenue under the Stability AI Community License. Organizations above $1M in annual revenue are directed to contact Stability for an Enterprise License. Hosted/API costs depend on the selected platform or route.
Source: Stability AI SD3.5 announcement
Model Snapshot
Stable Diffusion 3 is the legacy Stability baseline in the published examples. It helps show how Stability's output changed between SD3 and SD3.5 when interpreting the same keyword prompts.
- Best For
- SD3-era baseline results and comparisons against newer SD3.5 outputs
- Pricing Note
- Pricing depends on whether generation happens through Stability's API, self-hosted infrastructure, or another platform. Because Stable Diffusion 3 is included as a legacy comparison, the page should avoid presenting it as the current Stability pricing benchmark.
- Release Context
- Stable Diffusion 3 predates the SD3.5 family. SD3.5 should be framed as the later refinement path represented by the newer Stable Diffusion results.
- Deprecation / Legacy Note
- Stable Diffusion 3 remains a legacy comparison point beside the newer SD3.5 results.
*As of June 27, 2026: Stability pricing and licensing should be described by generation route, not as a single per-image price. For the current Stability section, SD3.5 Medium is the stronger pricing/licensing reference than legacy Stable Diffusion 3.
Source: Stability AI SD3.5 announcement