Pricing and limits
Sandboxes and storage are paid from the same budget as model requests. You top it up at pirol.ai/console/budget — there is no separate plan and no subscription. The prices are on the pricing page.
How you are billed
| What | Billed by |
|---|---|
| CPU | core-hour; one core is 2 vCPU |
| Memory | GiB-hour |
| Storage | GiB-month, backups and templates included |
| GPU | GPU-hour, on top of CPU and memory, see Training |
A sandbox reserves 0.125 cores and 128 MiB and may use more than that. You pay the higher of what it reserves and what it actually uses, by the second, for as long as it runs.
A stopped sandbox costs nothing. Its storage and its backup still do, because both are kept until you remove them.
Every workspace gets a free allowance each calendar month. It resets every month and does not accumulate. It needs a confirmed email address. All prices exclude VAT.
When your budget runs out
Two steps, so that an empty budget does not cost you your work:
- Running sandboxes are stopped ten minutes after your budget reaches zero. Nothing is
deleted, and
/sandboxis backed up to your storage first. Top up and start them again withnido start— they do not restart by themselves, because that would spend your new budget without asking. - Storage is deleted one day later — volumes, backups and templates. A sandbox that uses
one of those volumes is removed with it: without its storage it cannot start again. Sandboxes
without storage stay, and start once you top up. We only delete after we have emailed the
owners of the workspace with the date, and
nidoand the console show the same date. Top up before it and nothing is deleted.
Limits
| Limit | Today |
|---|---|
| Sandboxes per account | 100 |
| Storage per account | 10 volumes, 100 GB in total |
| Interactive shell | not yet; use nido exec |
| Backups and templates | 5 GB per backup, 20 templates; keep larger data in storage |
GPU sandboxes have limits of their own, see Training.