Although the next generation of Stable Diffusion, an AI model for text-to-image, is still in preview, Stability AI has already opened it to some developers through its API and new content creation platform. Stability AI announced that developers can now access Stable Diffusion 3 through its developer platform.

Stability AI revealed that they are collaborating with the API platform Fireworks AI to provide strong support for companies that wish to use both models simultaneously. They also plan to host model weights on their own servers in the near future through Stability AI's membership services.

In February, the company previewed Stable Diffusion 3 to a select group of developers. According to Stability AI, Stable Diffusion 3 performs "equally or better" than other text-to-image generators such as OpenAI's DALL-E 3 and Midjourney v6 in terms of "layout and prompt adherence." The model adopts an architecture called multimodal diffusion transformer, aiming to enhance text understanding and spelling abilities.

Stability AI also unveiled a new content creation platform called Stable Assistant Beta, which will provide Stable Diffusion 3 and other models. The company describes Stable Assistant Beta as a "friendly chatbot" that subscribers can use to access the latest models for generating images, writing content, or matching photos with text through conversation. Although Stable Assistant Beta is not yet available to the public, the company has invited a small number of users to try its early access version.

Despite the availability of Stable Diffusion 3 through the API and Stable Assistant Beta, Stability AI emphasizes that both models are still in the preview stage and not fully open to the public. The company states that they are expanding access to the models primarily for further improvement in collaboration with their community. At the same time, they have taken and will continue to take reasonable measures to prevent misuse of Stable Diffusion 3 by malicious actors.