Kavram is testing a production image generation API that uses a pipeline-split distributed GPU network to allow developers to ship open image models without managing hardware themselves.

  • The service provides one request shape for six image models, supporting queued, processing, and completed job states.
  • Integration paths are available via curl, Python, and Node.js.
  • Early pricing for FLUX.1 Schnell at 1 MP is $0.0012 per completed image, compared to fal.ai’s listed $0.003/MP.

The team is seeking technical feedback on the minimal reproducible artifacts expected before testing a new image inference API.