Modal Labs Quickstart: Serverless Python GPUs in Code (2026)

Sign up via GitHub and run `modal setup` in your local terminal.

• •Install the modal python package and run authentication.

pip install modal
modal setup

Define your container environment directly in Python with zero Dockerfiles.

• •Modal spins up an A100 or H100 in under 3 seconds, runs your code, and shuts down instantly.

import modal

app = modal.App("quickstart")

@app.function(gpu="A10G")
def square(x):
    import torch
    print(f"CUDA available: {torch.cuda.is_available()}")
    return x ** 2

@app.local_entrypoint()
def main():
    print(square.remote(42))