CreateOS Sandbox vs Cua Fleet
Pick Cua if you are building computer-use agents that need Windows or macOS, or want open-source tooling you can run locally or in your own AWS, Google Cloud or Modal account. Pick CreateOS Sandbox if a Linux desktop with Chrome is enough and you want published per-second prices, a micro-VM per sandbox, and pause and fork.
Pricing at a Glance
CreateOS rates verified 17 October 2026. Cua Fleet 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
- Compute
- $0.0362 /vCPU-hr
- Memory
- $0.0116 /GiB-hr
Cua Fleet
Cua publishes no rate for hosted Fleet compute; its pricing page lists Cua Spaces for your own machines at $0 (Free), $18 per month (Pro) and $30 per seat per month (Teams), as of 17 October 2026.
Where CreateOS Sandbox Fits Better
What CreateOS Sandbox brings to this matchup, with a plain note on where Cua Fleet has the edge.
Published per-second rates ($0.03616363 per vCPU-hour, $0.01159025 per GiB-hour); Cua does not publish a Fleet compute rate.
Every sandbox, desktop included, is a Firecracker micro-VM with its own guest kernel; Cua Fleet runs container images under gVisor.
The desktop:1 image ships XFCE, Google Chrome and a computer-use API, with pause and fork on the same sandbox.
Where Cua Fleet is strong
Cua is the most complete open-source computer-use stack in this set: Linux, Windows and macOS desktops, a driver for real machines, more than 100 supported vision-language models, SDKs in five languages, and one SDK that targets your laptop, Cua's cloud or your own cloud.
Which to Pick
Pick CreateOS Sandbox if
- A Linux desktop with Chrome covers your computer-use tasks.
- You need a published price before you start.
- You want to pause a desktop with its memory and fork it into copies.
Pick Cua Fleet if
- Your agent has to operate Windows or macOS applications.
- You want MIT-licensed tooling and the option to run desktops in your own AWS, GCP or Modal account.
- You want warm pools that autoscale from zero.
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 | Cua Fleet |
|---|---|---|
| Isolation | Firecracker micro-VM with its own guest kernel, root filesystem and network identity | Fleet pools run containers under gVisor and VMs under KubeVirt (amd64 only); no macOS on Fleet |
| Pause and resume | Saves disk, memory and CPU state; vCPU stops billing while paused, memory and storage still bill | Claims are leases renewed by TTL (15 minutes default); no snapshot API for Fleet sandboxes found |
| Fork | Fork a paused sandbox into many independent copies, each with its own id and IP | Not published |
| Outbound network default | Open with no rules; deny-by-default once any host, IP, CIDR or wildcard rule exists, enforced on the host outside the VM | Not published |
| Inbound access | Closed by default; opt-in public HTTPS URL, plus API-authenticated shell and port tunnels | Named services route authenticated clients to sandbox ports; service URLs can be signed |
| Concurrency | Free 1, Beginner 5, Pro 20, Enterprise 30; paused sandboxes count toward the cap | Managed pools autoscale from zero to 10 by default; your own pools set their bounds |
| Session length | No lifetime cap; optional idle auto-pause from 60 s to 24 h, off by default | Claim TTL 15 minutes by default, renewed while held; managed pools deleted after 30 minutes idle |
| Billing | Per second: $0.03616363/vCPU-hr, $0.01159025/GiB-hr, storage $0.0001584/GB-hr; no egress fees, 50 GiB bandwidth budget per sandbox | Fleet compute rate not published; warm replicas are billed |
| SDKs and interfaces | TypeScript, Go, Python, Rust, C#, Java (all but TypeScript pre-1.0), CLI, MCP, REST | Python, TypeScript, Swift, Kotlin and Rust SDKs, CLI, MCP server, Terraform provider |
| Self-host or BYOC | Enterprise only, onboarded by our team; not open source | Open source (MIT); run locally or in your own AWS, Google Cloud or Modal account |
Common Questions
Does Cua Fleet support Windows and macOS desktops?
Cua supports Linux, Windows and macOS overall, but macOS is not offered on Fleet; Fleet pools serve Linux and Windows. CreateOS Sandbox offers Linux only, through its desktop:1 image with XFCE, Google Chrome and a computer-use API.
How much does Cua Fleet compute cost?
Cua does not publish a Fleet compute rate; its docs say warm replicas are billed, and the pricing page now covers Cua Spaces for your own machines. CreateOS Sandbox bills desktops like any sandbox: $0.03616363 per vCPU-hour and $0.01159025 per GiB-hour, per second.
Can I snapshot a Cua Fleet sandbox and fork it?
No snapshot API for Fleet sandboxes is documented, and a released sandbox may be recycled, so results must be copied out. CreateOS Sandbox pauses a desktop with its disk, memory and CPU state, and forks the paused sandbox into independent copies.
Sources
Checked 17 October 2026.
See CreateOS Sandbox for yourself
Book a demo, or explore the Sandbox and its pricing.