Run Python compute workloads anywhere, with only a predict function. Get started

Bring your predict function

No need for nasty YAMLs or Dockerfiles. Just bring your code

# Our predictor says a greeting
# Create the predictor with the Function CLI

Provision on powerful GPUs

Run your predictors on powerful cloud A40 and A100 GPUs in addition to CPU instances.

Iterate with your team

Don't wait for a new CI build or Vercel deployment. Have your team mates /predict.

Deploy everywhere

And only pay for time spent making predictions—not time spent idle or on cold starts.