At Stability AI, we’re committed to releasing high-quality Generative AI models and sharing them generously with our community of innovators and media creators.  We acknowledge that our latest release, Stable Diffusion 3 Medium, didn’t meet our community’s high expectations, and recognize that the commercial license originally associated with SD3 caused some confusion and concern. We’ve made a few improvements to address these concerns. Read the full update here:http://stability.ai/news/license-update… (1/7)

Non-commercial use remains free.  People and organizations that install and run Stability AI models on their own devices for non-commercial use, can continue doing so for free.  This applies to most of our community, including researchers that usually publish their research for free, free open-source developers, full-time students, teachers, hobbyists, “fine-tuners” that improve our models and distribute them, etc. (2/7)

Free commercial use for individuals and small businesses. If you or your small business use Stability AI’s models under the ‘Stability AI Community License’ to create derivative products (e.g. finetunes of Stable Diffusion 3) or integrate our models within your product or service, it’s free as long as your annual revenues don't exceed USD $1M. (3/7)

Fewer limits. There will be no restrictions on the number of media files that can be created with the ‘Stability AI Community License Agreement’. (4/7)

Only commercial users need to self-report.  If you integrate Stability AI’s models or derived products into your own commercial products or services, and your annual revenues are below USD $1M, you don’t need to pay anything to Stability AI.  We do ask that you fill out a brief form, and indicate what models are you most interested in. Once your annual revenues exceed USD $1M, you'll need to contact Stability for an Enterprise license. (5/7)

Improving Model Quality. Before we released SD3 Medium, our initial testing indicated that it was, in most cases, a much better base model compared to SDXL in terms of prompt adherence, diversity, detail, and overall quality.  However, the community quickly identified some critical quality issues mainly related to body poses and words that were too rarely seen in the training set. (6/7)

Continuous Improvement. SD3 Medium is still a work in progress. We aim to release a much improved version in the coming weeks. We are all researching how to best use this new architecture and model. As we learn more about the best ways to use the model and to fine-tune it, we will share our findings as well as those of others In the community. (7/7)