Backprop makes it easy to deploy your Machine Learning models into a scalable, production-grade, API.


Uploading models

Getting your models ready for production is easy: just drag and drop a .zip, or use our library. In just a few minutes your model will be available to you via our API. We're also framework-agnostic, so whether you use PyTorch, Tensorflow, or SKLearn, you can train models however you prefer and still deploy through Backprop.

Sign up on our dashboard, and then check out our docs to find out how to upload your own model.

SOTA models

Backprop includes state-of-the-art models out of the box, ready to be used for a variety of tasks on text and images. We have models like

  • OpenAI's GPT-2
  • OpenAI's CLIP
  • Google's T5
  • Facebook's BART

And more, with others in the pipeline.

Try out some demos in this sandbox — we've selected a few models and tasks to show off here. To tweak parameters and try more models, sign up on our dashboard. Check out our demo of CLIP here.

We also have a library that makes it easy to finetune and upload models for your specific tasks.


There's no reason to pay for infrastructure that's sitting idle. As we operate on a serverless architecture, pricing is usage-based: you only get charged when your model is actually being used. And when usage is high, our API is highly scalable — since models run in an isolated environment, your users won't experience a slowdown.