Modal
Serverless compute platform for Python, where functions are decorated in code and run in containers on CPU or GPU with per-second billing.
modal.compricingstatustermsdocs
A decorator, then a GPU.
Modal turns a Python function into a remote container. Decorate it, describe the image and the hardware it needs in the same file, and the platform provisions the machine, runs the code and tears it down again — billed by the second, with nothing charged while idle. The same mechanism covers model inference, sandboxes, training runs, notebooks and batch processing.
Plans are a free starter tier with a monthly credit allowance and a small seat limit, a paid team tier with a larger allowance and unlimited seats, and a quoted enterprise tier. Compute is metered per second: CPU by the core, memory by the gibibyte, GPUs by the card.
Worth knowing
Python is the interface, not one option among several. There is no general web-application runtime and no long-lived server — this is a compute platform that happens to serve HTTP, and an ordinary website would be working against the grain of it.
Container and GPU concurrency are capped per plan, so scale is a commercial tier as much as a technical one.
Identity
- Ownership
- VC-backed
- Status
- Active
Regions and law
- Regions
- United States, Canada, Mexico, Ireland, United Kingdom, India, Japan, Australia1
- Headquarters
- United States
Sources
- 2026-08-01: Regions
modal.com/docs/guide/region-selection - 2026-08-02: Entry price, Typical ceiling, Currencies, Free tier, Pricing model
modal.com/pricing - 2026-08-02: Billed, Paid, Exit
modal.com/legal/terms
Last reviewed 2026-08-01.
Complete this record.