Is Modal free? What it actually costs in 2026
Modal's Starter plan is $0 with $30 of compute credit a month, three workspace seats and limited scheduled and web functions; Team is $250 a month with $100 of included compute. Everything is billed per second.
Starter gives $30 of usage a month free — about 7.6 H100 hours at $0.001097 a second — for three seats. Beyond it you pay per second: CPU $0.0000131 per core-second, memory $0.00000222 per GiB-second, GPUs from $0.000164 (T4) to $0.001972 (B300) a second.
What the free tier actually includes
- $30 of compute a month, then per-second billing
- 3 workspace seats
- Scheduled and web functions limited on Starter
Where the wall is
The thirty-first dollar, which on an H100 arrives after about eight hours. Modal's scale-to-zero means idle costs nothing; a training or inference job that runs continuously reaches the credit in a day and bills per second from there.
After that: Starter $0 plus usage; Team $250 a month with $100 of compute and unlimited seats; Enterprise custom. GPU per second: T4 $0.000164, L4 $0.000222, A10 $0.000306, L40S $0.000542, A100 40 GB $0.000583, A100 80 GB $0.000694, H100 $0.001097, H200 $0.001261, B200 $0.001736, B300 $0.001972 (an H100 is about $3.95 an hour).
Can you self-host Modal? No — cloud only
Modal is a serverless cloud. vLLM, BentoML and Ray Serve are the self-hosted serving stacks compared here; the per-chip rental pages show what the GPU underneath costs elsewhere.
What it actually takes to run one yourself
Real RAM against the documented minimum, monthly cost, setup time, and the one thing that catches people out.
When paying is still the right call
Stay for bursty or unpredictable traffic — scale-to-zero means an idle endpoint costs nothing, and no self-hosted setup matches that. Stay while you are still finding out whether the model is worth serving at all; paying twice the GPU rate for a week of iteration is trivial next to the time it saves. Stay if you have no Kubernetes and no desire to acquire it: vLLM on your own cluster is cheaper per hour and considerably more expensive in attention. And stay if your workload is genuinely spiky batch work, which is the shape serverless was built for.
What you are locked into
Higher than it looks, and it is worth naming precisely. The model weights and inference code are portable — that part is just Python and PyTorch. What is not portable is Modal's own layer: `@app.function`, image definitions, volumes, secrets and scheduling all live inside your source files, so leaving means unpicking the platform from the application rather than redeploying it elsewhere. Teams that keep the model logic in a plain module and confine Modal decorators to a thin entry point keep the exit cheap. Teams that do not, do not.
If you would rather not pay: vLLM
vLLM
The standard open inference engine — the thing under most serving platforms.
Free and open source. You pay only for the GPUs you rent or own.
What will the GPU cost? Compare every company’s hourly price
See all 4 Modal alternatives comparedCompare the free options head-to-head
Read from Modal Labs's own pricing and documentation on 2026-09-26 — 33 days ago. Vendors change prices without notice; confirm before you commit.