CapabilitiesBilling models
CPU billed only while working
An idle processor costs nothing or very little.
Short answer: 5 of 16 providers offer it fully: Cloudflare Sandbox, Runtime, Sprites, Tensorlake and Vercel Sandbox. Checked 29 October 2026.
Who offers it
Part of Billing models, which explains why it matters.
Fully: Cloudflare Sandbox, Runtime, Sprites, Tensorlake and Vercel Sandbox. Not at all: Agent 37, Blaxel, Boat, boxd, Daytona, E2B, InstaBox, Lizard, Modal, Morph and Sandboxer.
Memory and disk are still charged for the full size while it runs.
pricing-page “CPU usage is based on active usage only” Source
A completely idle machine still pays for one twentieth of a core.
docs-confirmed “the same sandbox costs $0.03125 an hour, for memory plus the default CPU floor of 50 millicores” Source
Memory is also charged on actual use, and a minimum of 256 MB applies while running.
pricing-page “Cumulative CPU usage measured by cpu.stat” Source
Memory and disk are charged on the full allocation while it runs.
docs-confirmed “CPU tracks active runtime execution, not allocated vCPU time is used.” Source
Memory is still charged for the whole time the sandbox runs, including while it waits.
pricing-page “Time spent waiting for I/O (such as network requests, database queries, or AI model calls) does not count toward Active CPU.” Source
Note: A running machine pays its full compute rate whether it is working or waiting.
inferred “An awake instance still pays compute even when its CPU is idle.” Source
Note: The meter runs on the memory you allocated for as long as the sandbox is active, busy or not.
docs-confirmed “You are charged for memory (based on how much you allocated at sandbox creation) and storage while a sandbox is in active mode.” Source
Note: A running machine costs the same whether it is working or waiting.
docs-confirmed “You pay per second, only while a sandbox runs.” Source
Note: The processor is charged at the machine's full size for as long as it runs.
inferred “vCPU is the one component billed on the machine's size: its full vCPU count while it runs” Source
Note: You pay for the cores and memory you reserved, not the ones the work used.
docs-confirmed “Sandboxes are billed for the resources reserved: vCPU, RAM, and disk” Source
Note: The bill follows the cores and memory you gave the machine, not what it used.
docs-confirmed “based on the vCPU and RAM allocated to it” Source
Note: A waiting machine costs the same as a busy one of the same size.
inferred “While a machine is running, you pay for the vCPUs and memory you gave it.” Source
Note: Waiting time is charged at the same rate as working time.
inferred “Running sandboxes are billed per second, including idle time.” Source
Note: A machine that reserves more than it uses pays for the reservation.
docs-confirmed “You pay for max(request, actual).” Source
Note: The rate follows the largest of cores, memory and disk, so spare capacity is paid for.
inferred “MCUs = max(vCPUs, ceil(RAM/4GiB), ceil(disk/16GiB))” Source
Note: The price depends on the size you picked, not on how hard it works.
docs-confirmed “Metering is per second, from the moment a sandbox is up until it is stopped.” Source
Related
Need several capabilities at once? Add more in the capability finder. Something wrong here? Use the report link at the bottom of the page.