API Offering: Stable Diffusion 2.0 with Accelerated Inference

Dec 16, 2022


Stochastic is excited to share that we have added a Stable Diffusion 2.0 API to our platform! With our accelerated inference, generating images is 2x faster. Sign-up now for a free $20 credit, enough to generate 4,000 images using Stable Diffusion 2.0!

Stable Diffusion is one of the newest Latent Diffusion Models, which use a pre-trained text encoder (OpenCLIP-ViT/H) to generate high quality images using text prompts. With the newest release of Stable Diffusion 2.0, there are a variety of new updates for model users.

Stable Diffusion Improvements

According to a Stable Diffusion 2.0 article release from Stabillity.ai, the main improvements to Stable Diffusion 2.0 are as follows:

“The Stable Diffusion 2.0 release includes robust text-to-image models trained using a brand new text encoder (OpenCLIP), developed by LAION with support from Stability AI, which greatly improves the quality of the generated images compared to earlier V1 releases. The text-to-image models in this release can generate images with default resolutions of both 512x512 pixels and 768x768 pixels.” - Stability.ai, November 2022

The higher image quality and better model robustness are reasons why users should try Stable Diffusion 2.0 over the original V1. Using a larger, more accurate model to generate the images you need is a no-brainer, especially when deploying Stable Diffusion can be done in just a few minutes.

Accelerated Stable Diffusion 2.0 API

As part of Stochastic’s model offerings, Stochastic hosts a variety of model APIs, and we are excited to include Stable Diffusion 2.0 as part of our API options for users. Stochastic’s Stable Diffusion 2.0 API joins other APIs that are available such as GPT-J and BLOOM 176B.

Through Stochastic’s Stable Diffusion 2.0 API, there are a variety of key benefits that users have access to:

  • Lower Image Costs: Image generation now costs ~$.05 per image, with a reduced latency of ~3.8s per image
  • Reduced Training Costs: End-to-end fine-tuning requires ~1h 15mins for Stable Diffusion V1 ($0.7) and ~2h for Stable Diffusion 2.0 ($1.1)
  • Customization with User Datasets: Users can upload new images of subjects they would like to see in future generated images, to which the Stable Diffusion API will adjust by using renditions of these uploaded images for new image generations
  • Faster Training: With Flash Attention, Stochastic has enabled accelerated model training for the Stable Diffusion API
  • Support for both Stable Diffusion Versions: For users who want the freedom to access both Stable Diffusion versions, API offerings for accelerated Stable Diffusion V1 and 2.0 are available


Stochastic is pleased to host the new Stable Diffusion 2.0 as an API to help users produce higher quality images. With accelerated inference, Stochastic’s Stable Diffusion 2.0 model has become even cheaper and faster for users!

Try it on our playground → Playground

To access the API → Sign-up

If you want to build your own → GitHub Repo