# Daytona vs Northflank: pricing, isolation and limits Both bill allocated CPU and memory; Northflank's rates are about a third of Daytona's, and it isolates in microVMs rather than containers. **Runtime runs the same job for 77% less than Daytona and 42% less than Northflank.** 1,000 one-minute runs of a 2 vCPU, 4 GiB sandbox cost $2.76 on Daytona, $1.11 on Northflank and $0.64 on Runtime, because Runtime bills the CPU your code uses rather than the CPUs it holds. A Runtime sandbox is one call, with no service or project to set up. Rates checked 23 September 2026. ## At a glance | | Daytona | Northflank | Runtime | | ------------- | ------------------------------------------------------- | ----------------------------------------------------------- | -------------------------------------------------------- | | Isolation | Containers by default; VM sandboxes as a separate class | Each workload in its own microVM, Kata Containers or gVisor | Firecracker microVM, own kernel, every sandbox | | CPU | $0.0504 per allocated vCPU-hour | $0.01667 per allocated vCPU-hour | $0.025 per vCPU-hour of measured CPU, with a small floor | | Memory | $0.0162 per GiB-hour | $0.00833 per GB-hour | $0.0075 per reserved GiB-hour | | Plan fee | None published | None published for compute | None; prepaid credit from $10 | | Free start | $200 of compute | 2 free services and 1 free database | 50 sandbox hours, no card | | Sandbox model | A sandbox from the SDK | A service with a deployment plan | A sandbox, created in one call | | Where it runs | US and EU | Northflank's regions, or your own AWS, GCP or Azure | One US region | | GPUs | GPU sandboxes; Windows machines too | L4, A100, H100 and others by the hour | CPUs only | ## Why is Northflank cheaper than Daytona? The two bill the same way: every vCPU and every GB the sandbox holds, for as long as it runs, busy or waiting. Only the rates differ. Northflank charges $0.01667 a vCPU-hour, a third of Daytona's $0.0504, and $0.00833 a GB-hour, about half of Daytona's $0.0162 per GiB-hour. So the same sandbox costs 60% less on Northflank at every level of use. Runtime counts differently. It bills measured CPU at $0.025 a vCPU-hour, with a floor of a twentieth of a vCPU, so an agent that waits on a model most of the minute pays for a small share of its CPUs. That makes Runtime the cheapest of the three for agent work. A job that keeps every CPU busy for the whole run is the exception: there Northflank's allocated rate is lower than Runtime's measured one. ## Cost for the same job 1,000 runs of a 2 vCPU, 4 GiB sandbox (4 GB on Northflank). Each run lasts 60 seconds and keeps the CPU busy for 20 CPU-seconds: an agent that mostly waits for a model. ``` Daytona CPU 1,000 × 60 s / 3,600 × 2 × $0.0504 = $1.68 Memory 1,000 × 60 s / 3,600 × 4 × $0.0162 = $1.08 Total $2.76 Northflank 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 $1.11 Runtime 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 $276.00 on Daytona, $111.10 on Northflank and $63.89 on Runtime. With both CPUs busy for the whole minute, it is $1.33 on Runtime against $2.76 on Daytona and $1.11 on Northflank, so for work that uses every CPU all the time Northflank costs less than Runtime. Disk, network, plan fees and free allowances are left out. ## What does it take to start a sandbox? - **Daytona:** `daytona.create()` from its SDK returns a sandbox. - **Northflank:** a sandbox is a service. You create it with a deployment plan, and Northflank runs it in its own microVM. - **Runtime:** `Sandbox.create()` returns a running sandbox, with no service, project or plan to set up. A default sandbox took 351 ms at the median from the create request to the first Python result, across 20 starts on 24 September 2026. ```ts import { Sandbox } from "withruntime"; const box = await Sandbox.create({ funding: "trial" }); try { console.log((await box.exec("python3 -c 'print(6 * 7)'", { check: true })).stdout); } finally { await box.stop(); } ``` ## Which one to pick - **Pick Daytona** for Windows machines, GPU sandboxes from the same SDK, or the $200 free start. - **Pick Northflank** for jobs that keep every CPU busy, GPUs by the hour, or sandboxes in your own AWS, GCP or Azure account beside databases and services on one platform. - **Pick Runtime** for agents that wait on models: a microVM per sandbox at under a quarter of Daytona's price and 42% under Northflank's, pause and [forks](/glossary/sandbox-fork) that keep memory and running processes, and one call to start. ## Switch from either one From Daytona, Runtime's SDK runs your code after one import change: ```ts no-run import { Daytona } from "withruntime/daytona"; // was: from "@daytona/sdk" ``` ```python no-run from withruntime.daytona import Daytona # was: from daytona import Daytona ``` From Northflank, replace the service with a sandbox: create, exec, files and stop each map to one call in the [migration table](/docs/migrate#map-the-calls). Or give your coding agent the one instruction in [migration](/docs/migrate): it switches the project on a branch, tests it on the free trial and reports the monthly saving. `npx withruntime compare --from northflank` prices the sandboxes you ran at Northflank's published rates, and `npx withruntime switch --from daytona` before your first top-up matches that top-up with credit, up to $100. The full side-by-sides are [Runtime vs Daytona](/docs/daytona-alternative) and [Runtime vs Northflank](/docs/northflank-alternative). ## Sources Daytona pricing and Northflank sandboxes pages rechecked 25 September 2026; the rest checked 23 September 2026. - [Northflank pricing](https://northflank.com/pricing), [Northflank sandboxes](https://northflank.com/product/sandboxes), [Sandboxes on Northflank](https://northflank.com/docs/v1/application/sandboxes/sandboxes-on-northflank) - [Daytona pricing](https://www.daytona.io/pricing), [Daytona sandboxes](https://www.daytona.io/docs/en/sandboxes) - Runtime [pricing](/docs/pricing) and [speed](/docs/speed) Facts on this page were checked on 25 September 2026.