Modal

ship

Serverless GPU and compute platform for running ML and AI workloads.

Visit modal.com →Add to kit Open in Noemium

ledger entry · observed by @whysanesanders

The easiest way to deploy a Python function onto a GPU without touching Kubernetes. Excellent for fine-tuning, inference, and batch jobs. Pricing is usage-based, so runaway jobs can get expensive fast.

Strengths

Easiest way to put a Python function on a GPU without touching Kubernetes.
Fits fine-tunes, inference and batch jobs that should not own a cluster.

Known limitations

Python-first; other languages have limited support.
Usage-based GPU costs require cost monitoring.
Cold starts can be noticeable for some workloads.

Use it when

Python ML jobs that spike and then go quiet.
Teams that want serverless GPUs more than a long-lived box.

Skip it when

The runtime is not Python.
Runaway GPU jobs would go unmonitored.
Cold starts would show up in the user-facing path.

Buy-check

·Price — freemium, free tier ($30/mo compute); pay-as-you-go beyond as of 2026-09-01. Open the pricing page, not a screenshot.
·Free tier — Free tier exists — burn it first.
·Cancel fallout — What happens to your data/projects on cancel — check the ToS, not the FAQ.
·Seat math — Per-seat pricing multiplies quietly. Count seats before the annual plan.
·Who skips it — Also skipped when: The runtime is not Python.; Runaway GPU jobs would go unmonitored..