Modal
Python-first serverless for ML, batch jobs, and long-running compute.
Modal is the most Python-idiomatic serverless platform on the market. You add a decorator, type `modal run`, and your function runs on Modal's infrastructure with the dependencies you declared in code. GPUs (A10, A100, H100) are first-class and priced per-second.
For an indie ML or data project, Modal removes the worst part of the stack — packaging Python and procuring GPUs. The $30/month free credit covers a non-trivial amount of inference. It's the wrong tool for a Rails app; right tool for "I want to run Whisper transcription for paying users this weekend."
Pros & cons
- + Run Python functions in the cloud with `modal.run`, no Dockerfile needed
- + GPU access (A10, A100, H100) priced per-second
- + Excellent cold-start performance even for big Python images
- + $30/month free credit is enough for many indie ML side projects
- − Python-only — no path for Node, Go, Rust, etc.
- − Pay-per-second model means estimating monthly cost requires actual usage data
- − Not the right shape for a typical Django web app
Plans & pricing
| Plan | Price | CPU | RAM | Disk | Bandwidth |
|---|---|---|---|---|---|
| Starter (free credits) $30/mo free credits | Free | — | 0 GB | — GB | 0 GB |
| CPU (per-second) | Free | $0.000111/s per vCPU | 0 GB | — GB | 0 GB |
| A10G GPU | Free | $0.000306/s | 24 GB | — GB | 0 GB |
Free tier: $30/month free compute credits
Features at a glance
Similar hosts
Scaleway
from $1.84/moFrench cloud with bare-metal ARM and a Heroku-style PaaS layer.
Vercel
from FreeThe default home for Next.js, with a polished DX and a love-it-or-hate-it bill.
Netlify
from FreeVercel's older sibling — broader framework support, more even-keeled pricing.
Cloudflare Pages
from FreeStatic + edge-functions hosting on the world's biggest CDN, free for almost everyone.