Pay for what you generate and run.

No subscription and no seats. A world is paid for once, and every run in it after that is free. A task on our machines pays for the seconds it ran. You see the price before anything starts.

Draft

$0.23 a world

230 provider credits

Explore and sweep.

  • Generated once, kept for you
  • Every later run in it free
  • 43 on the starting credit
The normal tier

Standard

$1.58 a world

1,580 provider credits

The normal tier, with a metric scale.

  • Generated once, kept for you
  • Every later run in it free
  • 6 on the starting credit

Plus

$3.08 a world

3,080 provider credits

Larger worlds, billed up to twice.

  • Generated once, kept for you
  • Every later run in it free
  • 3 on the starting credit

Reuse

$0.00 a run

0 credits

A world you already built.

  • Unlimited runs
  • Same scenario, prompt and model
  • Tomorrow, or next month

Nothing to reserve. There is no plan to pick and no minimum. An idle account costs nothing.

Priced before it runs. Every submission is quoted and checked against your balance first, so a run cannot cost more than you expected.

Charged for what ran. Cancel a sweep after ten of a thousand runs and you pay for ten. The rest is released.

Built once, run forever. A world is generated once and kept, filed under the scenario and prompt that built it. A thousand-scenario sweep costs one world, not a thousand, and running it again costs nothing at all. Credits never expire; top up from $5.00.

Compute by the second.

Every task on a managed pool - a world task, a twin validation, a scene from a photo, an evaluation, a DimOS eval case, a mutational sweep - pays for the time RunPod billed it, at the pool's rate. Your own machines cost nothing here.

CPU

$0.44 an hour

120 micro-dollars a second · runpod:cpu

Physics without a GPU: twin validations, evaluations, mutational sweeps.

  • The seconds RunPod bills, no more
  • Nothing while a pool is idle
  • Reserved before it runs, settled after

GPU

$2.16 an hour

600 micro-dollars a second

Every other managed pool: rendering views, world tasks, scenes from a photo.

  • The seconds RunPod bills, no more
  • Nothing while a pool is idle
  • Reserved before it runs, settled after

GPU sandbox

$2.16 an hour

600 micro-dollars a second · runpod:gpu

DimOS eval cases: your code, in a fresh container per task, on a GPU.

  • The seconds RunPod bills, no more
  • Nothing while a pool is idle
  • Reserved before it runs, settled after

Your machines

$0.00 an hour

a self-hosted pool

Your runner, your hardware, your simulator.

  • No compute charge from us
  • Worlds priced as above
  • Nothing connects in
Managed poolAn hourMicro-dollars a second
runpod:mutational$0.44120

Quoted from an estimate, charged for the time. A quote reserves a world task 10 min; a mutational sweep 30 min; a twin validation 5 min; a scene from a photo 15 min; an evaluation 15 s an episode, at least 1 min; a DimOS case 2 min an episode. A spec may ask for less with estimateSec; what is charged is always the time RunPod billed, and the rest of the reservation is released.

Model calls at cost, plus 25%. A scene from a photo asks a vision model what is in it and an image model to take it out. Those calls are charged at what the provider charged, plus 25%, and nothing else does that.

On a self-hosted pool the machines are yours, and so is their cost: SimCloud charges for the worlds a task generates and nothing for its compute. Mutational, the optional add-on, is a separate product priced on its own site.

Twin certification.

A certified twin of your robot in the public catalogue: one canonical spec, exports that agree, fitted to your robot's recordings, run in more than one engine, with a report card anybody can read and a badge for your README.

  • Your robot's twin, built and checked with you: from your URDF or MJCF and your LeRobot recordings to a version that passes, or a report that says what does not.
  • Re-checked daily while it is public, so the badge keeps meaning what it says.
  • Or validate your own: push a bundle from the console, pay for the validation's compute as above, and publish it when it is certified.

Priced per robot, with its recordings and our review.

simcloud@mutational.in