Self-Hosted Licenses

Use of the Core Models (including Stable Diffusion 3 2B) is free for everyone, unless you’re using the Core Models for a commercial purpose and you or your organization generate over USD $1M (or local currency equivalent) of annual revenue, regardless of the source of that revenue.

No, unless you’re making over USD $1M of annual revenue, regardless of the source of that revenue, then the Core Models and Derivative Works thereof are  free for you to use.

Any business making under USD $1M annually of revenue, regardless of the source of that revenue, doesn’t need to pay to use or distribute the Core Models.

No you won’t need to pay - create away!

The Stability AI Community License allows for research, non-commercial, and commercial use of the Core Modelsfor individuals or organizations that generate under $1M (or local currency equivalent) of annual revenue, regardless of the source of that revenue. 

So if you use Stability AI’s models for research purposes only, this falls within the Community License and you don’t need to pay, regardless of the amount of revenue generated by your organization. However, if a Core Model, or a Derivative Work thereof, is utilized in your research, and that research will be used for a commercial purpose (i.e., within a product or service), you must register with Stability AI and may also need to obtain a paid Enterprise License if the annual revenue of your organization, or the organization(s) performing the commercial use, exceeds $1M.

Congratulations for hitting this impressive milestone! Please contact Stability AI in order to purchase an Enterprise License.

Stability AI is focused on contributing to the “community” as well as establishing a sustainable business by serving enterprise customers. This allows us to continue investing in developing cutting edge AI technology for the benefit of everyone. Declining a new enterprise customer does not align with our goal to make our models available to as many people as possible.

No, LoRAs, hypernetworks, finetunes, retrain, retrain from scratch, etc, are not “foundational models”. A foundational model is a “new” model architecture that’s derivative of a Core Model.

LoRAs, hypernetworks, fine-tunes, retrains, etc, are not "foundational models". If you have created a LoRA, you can use its outputs to train other LoRA’s. You can't use a Core Model, or its outputs, to create a “new" model architecture.

Like most agreements, the Community License is revocable if you violate its terms. Otherwise, Stability AI cannot comply with its own legal obligations, promote the safe use of our technology, and protect the hard work and considerable resources that we have invested into developing AI tools for our community.

Yes, as between you and Stability, you own outputs generated from the Core Models or Derivative Works (such as fine-tunes) and therefore can use those outputs at your discretion, so long as you do so in compliance with applicable law and our AUP. 

We strive to develop leading open models. We employ some of the leading researchers in the world and they are working on amazing things. When we release our models, we ensure that they are at the top end of the relevant performance rankings, or that they bring additional options to the community in terms of form factors like size and speed.