Daytona vs Runloop: pricing, isolation and limits

Both bill allocated CPU and memory while a sandbox runs; the same job costs $2.76 on Daytona and $5.33 on a Runloop devbox.

Runtime costs 77% less than Daytona and 88% less than Runloop for an eval or agent job that waits on a model. 1,000 one-minute runs of a 2 vCPU, 4 GiB sandbox are $2.76 on Daytona, $5.33 on Runloop and $0.64 on Runtime, whose CPU is metered by use at $0.025 a vCPU-hour. Each Runtime sandbox is a Firecracker microVM, and each can pause with memory and running processes. Rates checked 23 September 2026.

At a glance

	Daytona	Runloop	Runtime
Isolation	Containers by default; VM sandboxes as a separate class	microVM, with a container inside	Firecracker microVM, own kernel, every sandbox
CPU	$0.0504 per allocated vCPU-hour	$0.108 per CPU-hour while the devbox runs	$0.025 per vCPU-hour of measured CPU, with a small floor
Memory	$0.0162 per GiB-hour	$0.0252 per GB-hour	$0.0075 per reserved GiB-hour
Disk	First 5 GiB free, then $0.000108 per GiB-hour	$0.00034236 per GB-hour	Included while running; paused storage $0.08 per GB-month
Plan fee	None published	Basic free; Pro $250 a month plus usage	None; prepaid credit from $10
Free start	$200 of compute	$50 of credit, no card; 3 running devboxes on the trial	50 sandbox hours, no card; 8 sandboxes at once
Pause and resume	Files and memory on VM sandboxes	Suspend keeps the disk, processes restart; on Pro	Files, memory and processes, every sandbox

Which is better for agent evals and benchmarks?

Runloop sells benchmarks as a product. Its Basic plan includes public benchmarks such as SWE-Bench, and Pro adds custom benchmarks built from your data. On Runtime you bring the harness and run each task in its own sandbox.

The cost of a large eval run is where they part. An eval task spends most of its time waiting for the model under test, and the three bill for that time at very different rates:

For 1,000 one-minute tasks	Runloop	Daytona	Runtime
20 CPU-seconds per task	$5.33	$2.76	$0.64
Both CPUs busy throughout	$5.33	$2.76	$1.33

A Runtime fork also copies a prepared sandbox with its memory and running processes, so a repository installed once can start several attempts from the same point. See agent evals and SWE-bench and RL environments.

Cost for the same job

Worked out for the first row: 1,000 tasks of 60 seconds on 2 vCPU and 4 GiB, 20 CPU-seconds of computing in each. Runloop's matching size is MEDIUM, with 2 CPUs, 4 GB and 8 GB of disk.

TextDaytona  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                                             $2.76Runloop  CPU    1,000 × 60 s / 3,600 × 2 × $0.108       = $3.60         Memory 1,000 × 60 s / 3,600 × 4 × $0.0252      = $1.68         Disk   1,000 × 60 s / 3,600 × 8 × $0.00034236 = $0.05         Total                                             $5.33Runtime  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

Scaled up to 100,000 runs a month, Daytona comes to $276.00, Runloop to $532.56 and Runtime to $63.89. With both CPUs busy for the whole minute, it is $1.33 on Runtime, $2.76 on Daytona and $5.33 on Runloop. Plan fees, storage while stopped, network and free credit are left out.

What survives between sessions?

A suspended Runloop devbox keeps its disk; compute and memory charges stop, and its processes must be restarted when it resumes. Suspend and resume are on the Pro plan. Daytona keeps files and memory when a VM sandbox stops. A paused Runtime sandbox keeps files, memory and running processes on every sandbox, for 1 to 365 days, and wakes on the next request:

Pythonfrom withruntime import Sandboxsbx = Sandbox.create(timeout_seconds=600)sbx.exec("echo state > /workspace/state.txt")sbx.pause()  # memory and files are kept; compute billing stopsagain = Sandbox.connect(sbx.id)again.wake(timeout_seconds=1200)again.extend(600)again.stop()

Which one to pick

• Daytona, for GPU sandboxes or Windows machines at a flat rate well below Runloop's, or a $200 free start.
• Runloop, for benchmarks, public and custom, delivered as a managed product, or for deployment into your own cloud account with the SOC 2, HIPAA and GDPR readiness it states.
• Runtime, for eval and agent runs at under a quarter of Daytona's price and an eighth of Runloop's, pause and forks that keep memory and processes on every sandbox, and no plan fee.

Moving to Runtime

Daytona code needs one import changed:

TypeScriptimport { Daytona } from "withruntime/daytona"; // was: from "@daytona/sdk"

Pythonfrom withruntime.daytona import Daytona  # was: from daytona import Daytona

Runloop code is ported through the migration table: a devbox is created with Sandbox.create(), commands run through exec, suspend and resume become pause() and wake(), and shutdown is stop(). npx withruntime compare --from runloop reprices your Runloop usage, and npx withruntime switch --from daytona before the first top-up matches it with credit, up to $100. A coding agent can do either port from the one instruction in migration, on a branch and on the free trial. See Runtime vs Daytona and Runtime vs Runloop.

Sources

Daytona and Runloop pricing pages rechecked 25 September 2026; the rest checked 23 September 2026.

• Runloop pricing, Devbox sizes, Devbox lifecycle, Runloop
• Daytona pricing, Daytona sandboxes
• Runtime pricing

Facts on this page were checked on 25 September 2026.