Skip to content

Pricing

One price list for everything. Cube and every Managed Application, Sandbox, Browser and Serverless Postgres, are billed by the Cube resources they use. Volumes are billed by their data and their NVMe cache, never for writes or reads.

MeterWhat is countedUnit price
Cube resources
CPUBusy CPU time, by the millisecond$0.01 / million CPU-ms$0.00001 / CPU-second$0.036 / CPU-hour$26.28 / CPU-month
MemoryHeld, by the millisecond, in 128 MiB steps$0.00222 / million GiB-ms$0.00000222 / GiB-second$0.008 / GiB-hour$5.84 / GiB-month
RequestsPer request into a Cube or through a sandbox link$0.30 / million
Traffic outNA · EUPer GB leaving North America and Europe$0.004 / GB
Traffic outAPACPer GB leaving Asia-Pacific$0.01 / GB
Kept asleep1What sleeping instances keep, and snapshots$0.00000761 / million GB-ms$0.00000000761 / GB-second$0.0000274 / GB-hour$0.02 / GB-month
Volumes
Data1All a volume holds, kept in tiered storage$0.00000761 / million GB-ms$0.00000000761 / GB-second$0.0000274 / GB-hour$0.02 / GB-month
NVMe cache1The part on the machine’s NVMe, the size you set; no write or read fees$0.0000228 / million GB-ms$0.0000000228 / GB-second$0.0000822 / GB-hour$0.06 / GB-month
Logs
Logs1What Cubes print that is kept, as printed$0.00000761 / million GB-ms$0.00000000761 / GB-second$0.0000274 / GB-hour$0.02 / GB-month

Plans

Start free with no card. Paid is $5 a month, with generous limits.

Free

No card needed

$0 / month

Each day

  • 1MCPU-ms
  • 6GiB-hours
  • 100Krequests
  • 3 GBtraffic out
  • 200 MBof logs

Kept at once

  • 15 GBof data
  • 15 GBon NVMe

Allows

  • 5 instances awake at once
  • 10 vCPUs and 10 GiB awake together
  • 20 sleeping instances kept
  • Instances up to 2 vCPUs, 4 GiB
  • Logs kept 3 days
  • A quota used up stops that until the next day

Paid

$5 / month

Each month

  • 60MCPU-ms
  • 250GiB-hours
  • 10Mrequests
  • 1 TBtraffic out
  • 10GB-months of logs

Kept each month

  • 20GB-months of data
  • 20GB-months on NVMe

Allows

  • 500 instances awake at once
  • 500 vCPUs and 1,000 GiB awake together
  • 10,000 sleeping instances kept
  • Instances up to 8 vCPUs, 16 GiB
  • Logs kept 30 days

Every limit of each plan

Every limit of each plan

Free’s quotas come back each day at 00:00 UTC; Paid’s inclusions each month, per meter. What is left is not carried over.

IncludesFree, each dayPaid, each month
Fee$0$5
CPUbusy CPU time, by the millisecond1M CPU-ms60M CPU-ms
MemoryGiB-hours awake6250
Requestsinto Cubes, and into sandboxes through their links100K10M
Traffic outto the internet, from any region3 GB1 TB
Data storedvolumes, sleeping instances and snapshots15 GB in all20 GB-months
NVMe cacheof volumes15 GB in all20 GB-months
Logswhat Cubes print, as printed200 MB printed10 GB-months kept
Past itThat stops until 00:00 UTCThe list price, from your balance
AllowsFreePaid
Largest instancea sandbox, a Cube’s instance, a browser, a database2 vCPUs, 4 GiB8 vCPUs, 16 GiB
Awake at onceinstances of every kind5500
Their vCPUs together10500
Their memory together10 GiB1,000 GiB
Sleeping instances kept2010,000
Instances madea minute10600
Instances of a Cubein a region, at most5500
Cubesper account10500
Variables and secretsper Cube64128
Volumesper project3, up to 10 GiB each500, up to 1 TiB each
All volumes togetherper project15 GiBNo limit
Secretsper project100, up to 64 KiB each1,000, up to 64 KiB each
Cubes’ logs kept3 days30 days
Log export destinations110
Readers following logsat once10100

A volume is at most 1 TiB on Paid, up to 16 TiB for an enterprise account by arrangement, and an instance mounts at most 8 that are not shared. We raise Paid’s limits on request. Every other limit is on the limits page.

Estimate a month

Pick what you run and how much: what a real workload uses of Cube’s resources, and what its month comes to on Paid and on Free.

JSON in, JSON out, a database call or two.

An instance awake 7.85 hours a day on average.

A 30-day month, this workload alone on the account. Sources

A month on Paid

$5.00

$0.35 a million requests

  • CPU26M CPU-ms$0.26
  • Memory29.4 GiB-h$0.24
  • Requests10M requests$3.00
  • Traffic out6 GB$0.02
Usage at the list prices
$3.52
Included in Paid
−$3.52
Paid’s fee
$5.00

Past what Free allows: requests.

Where the figures come from

Each use starts from a published workload where there is one: its machine, how long it runs, and how much of its CPU and memory is in use. What no source gives is our own guess, and says so. Every figure can be changed in the calculator.

Cube

An API’s and a website’s CPU is what our meter billed per request in our measurement, less per request as the volume grows. Cloudflare’s figures for the other two are for its Workers, which run in isolates; the same program here is billed more CPU, as the comparison with Workers below shows.

Serverless Postgres

The same program on Cloudflare Workers

Five programs, measured on both.2 A month of each at Workers Paid’s prices and at ours.3

Program1M a month10M a month100M a month
WorkersCubeWorkersCubeWorkersCube
Hello worldPlain text$5.00$5.00$5.00$5.00$32.00$32.00
A JSON APIHono and zod, a 1 KB order$5.00$5.00$5.00$5.00$32.00$32.58
Server-side renderingReact, a page of 48 products$5.00$5.00$5.00$5.00$33.83$46.894
A CPU-heavy handler1 MB of JSON parsed and hashed$5.00$5.00$5.15$5.07$38.08$36.45
A fetch proxyWaits 20 ms on an upstream$5.00$5.00$5.00$5.00$32.34$33.38

Compared with others

List prices for usage, before any plan’s fee, free allowance or discount. Each links to the page it was read from, with the day it was checked; none is more than 90 days old.

ProviderPriceRead from
SandboxAn hour of 2 vCPUs and 4 GiB, its CPU busy 10% of it
Glac$0.039 for the hour720,000 CPU-ms × $0.01 / M + 4 GiB-h × $0.008CPU only while busy; memory the whole hourOur list price
Vercel Sandbox$0.110 for the hour0.2 vCPU-h × $0.128 + 4 GB-h × $0.0212Active CPU; provisioned memory the whole hour (iad1)Vercel docs20 Oct 2026
E2B$0.166 for the hour2 vCPU-h × $0.0504 + 4 GiB-h × $0.0162$0.000014 a vCPU-second and $0.0000045 a GiB-second, while it runsE2B pricing20 Oct 2026
Daytona$0.166 for the hour2 vCPU-h × $0.0504 + 4 GiB-h × $0.0162By the second, while it runsDaytona pricing20 Oct 2026
Browser
Glac$0.034 an hour1 vCPU and 2 GiB, its CPU busy half the hour: CPU and memory at the list prices, from when Browser opensOur list price
Browser Use$0.02 an hourIn whole minutes, a minute at leastBrowser Use pricing20 Oct 2026
Browserbase$0.10–0.12 an hourPast the hours of its $99 or $20 planBrowserbase pricing20 Oct 2026
Volume
Glac$0.02–0.08 a GB-monthBy the bytes in use: $0.02 with none of it on NVMe, $0.08 with all of it; no write or read feesOur list price
Vercel Drives$0.05 a GB-monthBy the bytes in use; writes $0.004 a GB; every read $0.0015 a GB (iad1)Vercel docs20 Oct 2026
Fly Volumes$0.15 a GB-monthBy the size provisioned, attached or notFly.io docs20 Oct 2026

CPU, as StarSling measures it

StarSling’s open sandbox benchmark, its CPU score: Node.js web tooling at 4 vCPU and 8 GiB, in runs a second, higher is faster. These are the 13 providers of its run 36498291449, on its leaderboard on 2 October 2026; the Sandbox page’s 68% is Glac’s 29.0 over their median, 17.26. Glac is not in that run: we measured ours with StarSling’s suite on a production server in Coventry with no other sandbox busy, on 10 October 2026.

StarSling sandbox benchmarks, run 36498291449, read 2 Oct 2026

ProviderRuns a second
Glac measured by us29.00
boat.dev25.46
Daytona (VM)21.56
Microsandbox Cloud21.16
Namespace20.55
Novita19.55
Blaxel19.35
tamathe median17.26
Modal (VM)15.53
run.cloud14.89
Vercel13.01
Modal (gVisor)12.65
E2B12.60
Runloop8.22

How billing works

Questions about a bill: support@glac.ai

What counts as usage?

Cube and every Managed Application are metered alike: a sandbox, a browser session and a database each run as instances of Cube, and each is billed by what its instances use.

  • CPU: the CPU time an instance’s machine uses while it runs, as the host counts it: its vCPUs and the device threads working for it, to the microsecond. Time the host spends waiting on the machine’s behalf is not counted. A sleeping instance uses none.
  • Memory: the memory the host holds for an instance while it is awake, and while it sleeps paused in the host’s memory, in steps of 128 MiB, by the second. Pages it shares with its template and has only read are not counted. An instance that sleeps any other way holds none.
  • Requests: each request through the platform’s front door that a machine takes: at a Cube’s address or your own domain, at a sandbox’s preview links, and each SSH session.
  • Traffic out: the bytes instances send to the internet, answers to requests included, and logs exported to your OpenTelemetry destinations, priced by the region of the machine they leave from.
  • Stored: what a sleeping instance keeps (its memory image and the part of its disk that is its own; the template it was made from is free), snapshots, a volume’s data (the bytes in use, hour by hour, all of it kept in tiered storage) and what its backups take, by the GB-month.
  • NVMe cache: the part of a volume kept on the machine’s NVMe, the size you set, from none (read at the network’s speed) to all of it, by the GB-month. Nothing is charged for writing or reading a volume.
  • Logs: the bytes of what your Cubes’ instances print that is kept, as printed (not as stored, compressed), for as long as it is kept, by the GB-month. The lines for requests and for what the platform did cost nothing.

GET /v1/usage (or glac usage) gives these quantities by hour, day, instance or label, with the prices to multiply them by.

What happens when a quota or an inclusion runs out?

On Free, a day’s quota used up stops that one thing until 00:00 UTC: instances go to sleep when CPU or memory runs out, requests are answered with 429, and traffic to the internet stops. Free keeps at most 15 GB of data in all, as much of it on NVMe. Nothing is ever charged, and a Free account never owes.

On Paid, usage past a meter’s inclusion is charged at the list price and taken from your balance. When the balance is empty too, that meter stops:

  • CPU or memory: every instance of the account goes to sleep, with its data, and nothing new is made until it can be paid for.
  • Requests: requests to your instances are refused at the edges; the instances keep running.
  • Traffic out: traffic to the internet stops, and so do requests to your instances, since every answer is traffic out.
  • Logs: what your Cubes print from then on is not kept for the long term (it can still be read on the machine for a while); nothing else stops.
  • What is stored (sleeping instances, snapshots, volumes and their cache, and backups) cannot stop, and nothing of it is deleted for this: it is still there, and its cost becomes a debt.

Everything starts again as soon as Credits arrive. If you have no balance and no auto reload, we email you when a meter reaches 80% and 100% of its inclusion.

How do I move between Free and Paid?

Every account starts on Free, with no card. Moving to Paid takes a card: the $5 fee is charged at once, and the month’s inclusions start then. Moving back to Free happens when the month paid for ends; nothing is refunded, and what is left of the inclusions is not carried over.

How do top-ups and auto reload work?

On Paid, your balance is in Credits: one Credit pays for one dollar of usage past what the plan includes. A top-up is at least $20, charged to your card; the Credits arrive when the payment is collected. The Credits of a top-up last 12 months from that top-up, and we write 30, 7 and 1 days before they lapse.

Auto reload is off until you turn it on, and needs a card on record: when the balance falls below 5 Credits it tops up $20 (both numbers are yours to change). A new account may be charged at most $100 a month by auto reload and $500 a month by all top-ups; these rise after 30 and 90 days of payments. Credits cannot be withdrawn or moved to another account, and the plan’s fee is charged to the card, not taken from the balance.

What is deleted when an account owes?

Only a Paid account can owe, and only what is stored can take its balance below zero: sleeping instances, snapshots, volumes and backups. Less than $0.50 owed waits for the next top-up or subscription payment. From $0.50 the account is in debt: until it pays it can only read its bill, pay, and list, read and delete what it stores (or make a volume from a backup), and its instances sleep. We email when the debt starts and again on days 7, 14, 21, 27 and 29, and charge the card on record on the billing day.

If it still owes 30 days after the first email was delivered, everything the account stores is deleted: sandboxes and other instances, snapshots, volumes, backups, its own templates, and Cubes with their versions; the debt is then written off. The account itself stays, with its projects, keys, secrets, SSH keys, webhooks, domain registrations, events, audit log, ledger and bills. Deleting what you store stops a debt from growing.

Can I get a refund?

Credits are not refunded, and Paid’s fee is not refunded when you cancel. We refund a payment only if it was charged in error or the law requires it, for example when a consumer withdraws under section 21.4 of the Terms of Service; nothing there limits a refund the law gives you.

Are prices in US dollars?

Yes. We are not registered for tax in any country yet (GLAC PTY LTD is not registered for GST), so no tax is added to what your card is charged, and Credits are the amount you pay.