April 25, 2026
Serverless expands to image and video generation models, plus reliability improvements to auto-scaling.
Major features
Section titled “Major features”Image and video generation
Section titled “Image and video generation”Serverless now supports image and video generation models (text-to-image, text-to-video, image-to-image, image-to-video), with a matching Playground and job history experience.
Improvements
Section titled “Improvements”- Dedicated serverless deployments now each get their own unique subdomain endpoint.
- Improved reliability of serverless auto-scaling and external buffer capacity, reducing scaling and deployment failures.
- Fixed a bug where serverless pods were double-counted against workspace resource quotas.