The Stable Diffusion 3 and Stable Diffusion 3 Turbo models were released in preview in February. Now, Stability AI finally offers an artificial intelligence (AI) text-to-image model for some users. The company will let developers access AI models through the Stability AI Developer Platform API. It has partnered with API platform Fireworks AI to bring these models to the public. Notably, the AI ​​company’s next-generation AI image model features improved text understanding and spelling capabilities.

Stability AI announces limited availability of artificial intelligence models via a channel postal in its newsroom, and said: “As revealed in the Stable Diffusion 3 research paper, the model is equal to or better than state-of-the-art text-to-image generation systems such as DALL-E 3 and Midjourney v6 in terms of typography Rapid compliance based on human preference assessment.”

The new text-to-image model has two notable upgrades. First, its understanding of prompt text improved. It now better understands contextual knowledge in prompts and can generate images that more closely match user expectations. It also has improved spelling functionality. This is helpful when the user wants to generate an image with text in it. The company had earlier highlighted that AI will scrutinize what is being written and provide better output. Overall image quality is also expected to be improved.

These new AI models will also be open sourced in the near future, at least to some extent. The company said it will soon make model weights available for self-hosting through a Stability AI membership. Stability AI also explains that its model uses a new multi-modal diffusion transformer (MMDiT) architecture.

In addition to the AI ​​image generator, Stability AI invites a limited number of users to participate in early participation release A stable assistant currently in beta. The AI ​​assistant is powered by Stable Diffusion 3 and Stable LM 2 12B, the latter of which adds conversational capabilities. It can generate images from conversations, generate content, and improve the content to match the generated images. Currently, it is unclear when the company will release the new AI image model to all members.