Run · instant compute

Spin it up.
Throw it away.

Run lets anyone launch a disposable container or hop on a shared VM in seconds — no provisioning, no commitment. Perfect for experiments, notebooks, and quick inference. When you're done, it vanishes. You only pay for what you use.

<5scold start
$0idle cost
Per-secbilling
$ superrails run start \
    --image pytorch --gpu 1
 container up in 3.2s
 jupyter: run-x9.superrails.io
$ # run your experiment...
$ exit
 torn down · billed 4m12s · $0 idle
// own the hardware

Own a supercomputer.

Popular tiers from the store — built on Nvidia, hosted on the rails, rented to developers like you.

View all 10
// co-own

Or split the cost.

Join a funding pool from $500 and share the rental revenue, pro-rata, with other members.

Browse all pools
// frictionless by design

Compute when you want it. Gone when you don't.

For the moments between big infrastructure decisions — the quick test, the demo, the one-off run.

Disposable containers

Launch a fresh, isolated container from any image in seconds. Run your job, then let it tear down automatically.

Shared VMs

Hop onto a ready-to-go shared VM for lightweight work — no setup, no waiting, an environment that's already warm.

Per-second billing

Pay only for the seconds you actually run. Idle costs nothing — there's nothing left running to bill.

Notebooks & CLI

Open a hosted Jupyter notebook or drive it from the CLI/API. Bring your own image or use our prebuilt ML stacks.

GPU on tap

Attach a GPU to any container for quick inference or a fast experiment, then release it the moment you stop.

Shared rails

Runs on the same supercomputers owned via Rails — so spare capacity is always nearby and cheap.

// made for

When you just need
compute, right now.

Run is for the quick test, the demo, the one-off job. No accounts to provision, no clusters to babysit — just an environment that appears, does the work, and disappears.

  • Prototyping a model or running a notebook
  • One-off inference or batch jobs
  • CI runs and reproducible experiments
  • Workshops, demos, and teaching environments
run vs. cloud
Best forShort, bursty work
LifespanSeconds to hours
IsolationContainer / shared
Idle cost$0
Need long-lived?Use Cloud →
// find your fit

Which machine is right for you?

Set your budget and goal — we'll recommend the SuperRails tier that fits.

$100,000

Recommendation is illustrative — based on the same figures shown across the store. Not financial advice.