The four meters of per-second billing
- Running: its size’s hourly price, by the second. Each run is rounded up to a whole second and counts 10 seconds at least, unless the server ended it: out of memory or disk, a restart of its machine, an unpaid workspace, a new size.
- Paused: 30% of that price, by the whole second, with no minimum.
- Stopped: $0.03 per GiB a month of what its disk holds of its own in object storage, never empty space and never the template’s files.
- Volumes: $0.20 per GiB a month of their whole size.
Templates are not billed: your plan bounds how much they hold. The disk of a running or paused sandbox is part of its size’s price, never billed apart. A second counts at one size only, and pausing and waking never turns running time into cheaper time, nor counts a second twice.
| Size | vCPUs · memory · disk | Running, an hour | Paused, an hour |
|---|---|---|---|
| SMALL | 2 · 4 GiB · 10 GiB | $0.018 | $0.0054 |
| DEFAULT | 4 · 8 GiB · 20 GiB | $0.036 | $0.0108 |
| LARGE | 8 · 16 GiB · 40 GiB | $0.072 | $0.0216 |
| XLARGE | 16 · 32 GiB · 80 GiB | $0.144 | $0.0432 |
A SMALL sandbox running costs $0.000005 a second, and the shortest charge there is, a run of 10 seconds, costs $0.00005.
How the seconds are counted
The sandbox’s own server counts its seconds, as each change happens, from the moment its VM’s vCPUs are let run; the panel prices them to a billionth of a dollar, every minute for running and paused sandboxes and every hour for stored disks and volumes. get_sandbox shows a sandbox’s seconds and what it cost so far, in usage, and list_usage gives a row per UTC day, meter and size.
from pethost.sandboxes import Sandboxes, SandboxSettings, Size
sandboxes = Sandboxes() # reads PETHOST_TOKEN
sandbox = sandboxes.create_sandbox("base", settings=SandboxSettings(size=Size.SMALL))
sandboxes.run_command(sandbox.sandbox_id, command="python3 -c 'print(6 * 7)'")
sandboxes.stop_sandbox(sandbox.sandbox_id) # its run ends here: 10 seconds at least
usage = sandboxes.get_sandbox(sandbox.sandbox_id).usage
print(usage.running_seconds, usage.paused_seconds, usage.cost_micros) # cost in millionths
# The workspace's last 30 days: a row per UTC day, meter and size.
for row in sandboxes.list_usage().rows:
print(row.day, row.meter.name, row.size.name, round(row.unit_hours, 3), row.cost_micros)pethost sandbox new --name demo --size small
pethost sandbox run demo -- python3 -c 'print(6 * 7)'
pethost sandbox stop demo # its run ends here: 10 seconds at least
pethost sandbox show demo # its state, size and usage
pethost sandbox workspace # the plan's limits and what is usedStarts are free
Creating a sandbox, forking one (each copy) and starting a stopped one cost nothing. Your plan bounds how many come in any minute, hour and day: 36, 360 and 1,200 on the $20 plan, up to 270, 3,600 and 14,400 on the $2,000 plan. A start past a bound is refused (RESOURCE_EXHAUSTED with LimitReached), saying which window is full and when the call fits again. Waking a paused sandbox is never a start.
Plans and the balance
A plan’s monthly price is usage it includes: $20, $100, $500 or $2,000 a month, plus VAT. The $20 plan’s usage is 1,111 hours of a SMALL sandbox running, or 3,704 paused. Past it your balance pays, topped up with $10 to $1,000 at a time, which never expires. With both spent, running and paused sandboxes are stopped and nothing new starts, but nothing is deleted, and stored disks and volumes go on counting; a payment lifts it within seconds. It all begins with a trial: 7 days with $1 of usage and 4 vCPUs, sizes up to DEFAULT, a card required.
Two months, worked
An agent with one SMALL sandbox, working 2 hours a day and paused for the other 22: 60 hours running at $0.018 and 660 paused at $0.0054 make $4.64 of usage in a 30-day month, so the $20 plan covers about four such agents.
An eval fleet of 1,000 SMALL sandboxes a day, 60 seconds each: 500 sandbox-hours, $9.00 of usage a month, with no start fees, inside the $20 plan, whose 20 vCPUs run 10 SMALL sandboxes at once. The billing docs work through more cases.
How it compares
For 2 vCPU and 4 GiB, as their pages read on 21 November 2026: E2B bills $0.1656 an hour, by the second. boat.dev bills $0.018, as we do, and $0.03 to $0.07 a start past the starts its plan includes. Vercel Sandbox bills memory in 1-minute minimum increments and CPU only while it is busy, and Modal bills the larger of what a sandbox requested and what it used.