Fly Machines vs Northflank: pricing, isolation and limits

Both bill the CPU and memory a VM holds while it runs; 2 vCPU and 4 GB costs $0.0861 an hour on Fly and $0.0667 on Northflank.

For an agent that mostly waits, Runtime runs the same job for 55% less than a Fly Machine and 42% less than Northflank. 1,000 one-minute runs of a 2 vCPU, 4 GiB sandbox cost $1.44 on Fly Machines, $1.11 on Northflank and $0.64 on Runtime, because Runtime bills the CPU your code uses rather than the CPUs it holds. Rates checked 23 September 2026.

At a glance

Fly Machines are priced at the Ashburn (iad) rates. Northflank's isolation and regions are from its sandbox page, read 25 September 2026.

	Fly Machines	Northflank	Runtime
Isolation	Firecracker microVM	A microVM with Kata Containers, or gVisor	Firecracker microVM, own kernel
CPU	Inside a fixed size, billed while started	$0.01667 per allocated vCPU-hour	$0.025 per vCPU-hour of measured CPU, with a small floor
Memory	Inside the size; more at $5 per GB a month	$0.00833 per GB-hour	$0.0075 per reserved GiB-hour
2 vCPU, 4 GB	performance-2x, $0.0861 an hour	$0.0667 an hour	$0.08 an hour fully busy, $0.03125 waiting
Where it runs	18 Fly regions	5 Northflank regions, or your AWS, GCP, Azure or Oracle	Runtime's dedicated servers
Plan fee	None; pay as you go	None published for compute	None; prepaid credit from $10
Free start	Not stated on the pricing page	2 free services and 1 free database	50 sandbox hours, no card
Sandbox model	A Machine, created through the Machines API	A service with a deployment plan	A sandbox, created in one call

Which is cheaper for a machine that is always busy?

Northflank. Its allocated rates put 2 vCPUs and 4 GB at $0.0667 an hour (2 × $0.01667 plus 4 × $0.00833), against $0.0861 for a Fly performance-2x with 4 GB. It is also below Runtime's $0.08 an hour for the same size with both CPUs working the whole time. For a build farm or a batch job that keeps every core busy, Northflank's rate card is the lowest of the three.

The order changes as soon as the work waits. Fly and Northflank both bill a held vCPU the same whether it computes or idles on a model's reply. Runtime bills measured CPU, so the same sandbox waiting costs $0.03125 an hour: memory plus a floor of a twentieth of a vCPU. Most agent runs spend most of their time waiting.

A Fly performance vCPU is a whole core kept for the Machine. Runtime's default CPU is shared, and cpu: "reserved" keeps every vCPU for you when a job needs it.

Cost for the same job

1,000 runs of a 2 vCPU, 4 GiB sandbox (4 GB on Fly and Northflank). Each run lasts 60 seconds and keeps the CPU busy for 20 CPU-seconds: an agent that mostly waits for a model.

TextMachines    Size   1,000 × 60 s / 3,600 × $0.0861        = $1.44Northflank  CPU    1,000 × 60 s / 3,600 × 2 × $0.01667   = $0.556            Memory 1,000 × 60 s / 3,600 × 4 × $0.00833   = $0.555            Total                                           $1.11Runtime     CPU    1,000 × 20 s / 3,600 × $0.025         = $0.14            Memory 1,000 × 60 s / 3,600 × 4 × $0.0075    = $0.50            Total                                           $0.64

At 100,000 runs a month that is $143.50 on Fly Machines, $111.10 on Northflank and $63.89 on Runtime. With both CPUs busy for the whole minute, it is $1.11 on Northflank, $1.33 on Runtime and $1.44 on Machines: Northflank costs less than Runtime for work that never waits. Plan fees, storage, network and free allowances are left out.

Where can each one run?

Fly Machines run in 18 regions on Fly's own hardware, which suits an app that serves users worldwide. Northflank lists US West, US Central, US East, EU West and Asia East, and it can run its sandboxes inside your own VPC on AWS, GCP, Azure or Oracle, with GPUs up to H100 in the same platform. Runtime runs every sandbox on dedicated servers it operates, each in a Firecracker microVM with its own Linux kernel; see security and Kata Containers vs Firecracker.

Which one to pick

• Pick Fly Machines for an app or service that should run near its users, with Fly Volumes and Managed Postgres beside it.
• Pick Northflank for jobs that keep every CPU busy, for GPUs, or to run sandboxes in your own cloud account.
• Pick Runtime for agents that wait on a model: measured CPU, pause that keeps memory and processes for 1 to 365 days, forks of a running sandbox, and one call to create a sandbox, with no service or deployment plan.

Switch from either one

A sandbox is one call, with no app, service or deployment to set up first:

Pythonfrom withruntime import Sandboxwith Sandbox.create(funding="trial") as sbx:    run = sbx.exec(["python3", "-c", "print(6 * 7)"], timeout_ms=120_000)    print(run.exit_code, run.stdout)

Or give your coding agent the one instruction in migration: it switches the project on a branch, tests it on the free trial and reports the monthly saving. npx withruntime compare --from fly-machines (or --from northflank) prices what you ran at the old provider's rates, and npx withruntime switch --from fly before your first top-up matches that top-up with credit, up to $100. The full side-by-sides are Runtime vs Fly.io and Runtime vs Northflank.

Sources

Rates checked 23 September 2026; the Northflank sandbox page read 25 September 2026.

• Fly.io pricing, Machine CPU performance, Fly.io regions
• Northflank pricing, Northflank sandboxes, Sandboxes on Northflank
• Runtime pricing and security

Facts on this page were checked on 25 September 2026.