CreateOS Sandbox vs Google Cloud Run
Use Cloud Run to host your own stateless containers, APIs, batch jobs or GPU inference, with autoscaling and scale to zero. Use CreateOS Sandbox when each user or agent needs its own long-lived machine with a persistent filesystem, created by API, that can be paused, resumed or forked. Cloud Run's filesystem is in memory and gone when the instance stops.
Pricing at a Glance
CreateOS rates verified 17 October 2026. Google Cloud Run pricing as published on 17 October 2026. Rates billed per GB rather than per GiB, or as bundled sizes, are quoted as published rather than converted.
CreateOS Sandbox
Lower compute rate- Compute
- $0.0362 /vCPU-hr
- Memory
- $0.0116 /GiB-hr
Google Cloud Run
- Compute
- $0.0648 /vCPU-hr
- Memory
- $0.0072 /GiB-hr
Google lists us-central1 instance-based billing at $0.000018/vCPU-second and $0.000002/GiB-second, and request-based CPU at $0.000024/vCPU-second active, rounded up to the nearest 100 ms, as of 17 October 2026.
Where CreateOS Sandbox Fits Better
What CreateOS Sandbox brings to this matchup, with a plain note on where Google Cloud Run has the edge.
Sessions with no time cap; Cloud Run service requests end at 60 minutes at most.
Disk and memory survive pause, while Cloud Run's in-memory filesystem is lost when an instance stops.
Egress allowlists on the sandbox itself, where Cloud Run filtering means routing traffic through a VPC with firewall rules.
Per-user environments come from an API call, not from deploying a revision or job.
Where Google Cloud Run is strong
Scale-to-zero container hosting with per-100 ms billing, a large free tier and autoscaling, backed by Google Cloud networking, IAM and compliance scope, plus serverless L4 and RTX PRO 6000 GPUs without reservations.
Which to Pick
Pick CreateOS Sandbox if
- Each agent or user needs a private machine whose files persist across steps.
- Work must pause overnight and resume where it left off.
- You want to fork a prepared environment for parallel tries.
Pick Google Cloud Run if
- You deploy your own stateless web services or APIs and want autoscaling to many instances.
- You run batch job tasks of up to 7 days each.
- You need serverless GPU inference on L4 or RTX PRO 6000.
Head-to-Head Comparison
Every dimension side by side, checked on 17 October 2026 against each vendor's own docs and pricing pages. "Not published" means we could not find it on the vendor's pages, not that the feature is missing.
| Dimension | CreateOS Sandbox | Google Cloud Run |
|---|---|---|
| Isolation | Firecracker micro-VM with its own guest kernel | gVisor (gen1) or microVM (gen2); by default Cloud Run picks one |
| Create and destroy by API | Yes: REST API, CLI, MCP server and SDKs for TypeScript, Go, Python, Rust, C# and Java (all but TypeScript pre-1.0) | Admin REST API, gcloud, client libraries, Terraform; environments are deployed revisions or jobs |
| Memory snapshot (pause and resume) | Yes: disk, memory and CPU state saved; vCPU stops billing, memory and storage still bill | Not documented; in-memory filesystem lost when an instance stops |
| Fork | Yes: fork a paused sandbox into many independent sandboxes | Not published |
| Outbound network controls | Open with no rules; deny-by-default once any host, IP, CIDR or wildcard rule exists, enforced on the host outside the VM | Internet open by default; Direct VPC egress for firewall rules and network tags |
| Billing granularity | Per second while running: $0.03616363/vCPU-hr, $0.01159025/GiB-hr, storage $0.0001584/GB-hr | Per second, rounded up to the nearest 100 ms |
| Largest size | 8 vCPU / 16 GB on Enterprise; 8 vCPU / 8 GB on Pro | 8 vCPU / 32 GiB per instance (quotas page); GPU instances higher |
| GPU | None; no GPU shapes | Yes: L4 and RTX PRO 6000 Blackwell, one per instance |
Common Questions
How long can a Cloud Run request run?
Up to 60 minutes. The service request timeout defaults to 5 minutes and can be raised to 60, after which Cloud Run returns HTTP 504, while job tasks run up to 168 hours. A CreateOS sandbox has no lifetime limit and runs until you destroy or pause it.
Does Cloud Run keep files between requests?
Not reliably. Cloud Run's container filesystem is in memory and is lost when the instance stops, so durable state goes to mounted volumes or external storage. A CreateOS sandbox keeps its disk for its whole life, and pause also saves memory and CPU state.
Is Cloud Run cheaper than CreateOS for always-on compute?
On memory, yes: Cloud Run instance-based billing works out to $0.0072 per GiB-hour against $0.01159025 on CreateOS. On vCPU, CreateOS is lower at $0.03616363 against $0.0648. Cloud Run's monthly free tier of up to 240,000 vCPU-seconds also helps small workloads.
Sources
Checked 17 October 2026.
- Google Cloud Run: pricing, free tier, GPU prices, egress note, regions/tiers
- Google Cloud Run: isolation (gen1 gVisor, gen2 microVM)
- Google Cloud Run: max shape, instances, concurrency, job limits, startup timeout
- Google Cloud Run: service request timeout
- Google Cloud Run: job task timeout
- Google Cloud Run: in-memory filesystem, shutdown semantics
- Google Cloud Run: per-instance concurrency
- Google Cloud Run: GPU types, start time, min CPU/memory
- Google Cloud Run: ingress
- Google Cloud Run: IAM / public access
- Google Cloud Run: egress routing
- Google Cloud Run: compliance
- CreateOS Sandbox limits
- CreateOS pause, resume and fork
- CreateOS Sandbox pricing
See CreateOS Sandbox for yourself
Book a demo, or explore the Sandbox and its pricing.