# GPU agent sandboxes: prices at 2 vCPU / 4 GB (Nov 2026)

GPU sizes offered. 6 providers: one busy 1-hour session at 2 vCPU / 4 GB costs ≥ $0.0288 (Koyeb) to ≥ $0.238 (Modal).

1 h busy: all CPU billed. 1 h at 10% CPU: lower only where idle CPU isn't billed. Fleet month: 50 × 8 h × 22 days, busy, on the lowest plan that runs 50 at once (its fee, less the usage it includes). At once and Starts are on the terms priced. Fixed sizes use the smallest size that fits. "≥" means a billing term is not published. Method: https://roost.guide/methodology/

| # | Provider | 1 h busy | 1 h at 10% CPU | Fleet month | Pay | At once | Starts | Size priced | Billed on | Caveats |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Koyeb | ≥ $0.0288 [1] | ≥ $0.0288 | ≥ $253.44 | Pay as you go | ? | ? | eco-large (2 vCPU / 4 GB) | Running time | none |
| 2 | Northflank | ≥ $0.0667 [2] | ≥ $0.0667 | ≥ $586.61 | Pay as you go | 500 | ? | 2 vCPU / 4 GB | Running time | none |
| 3 | Daytona | ≥ $0.166 [3] | ≥ $0.166 | ≥ $1,457 | Pay as you go | ? | 300/min, then refused | 2 vCPU / 4 GB | Running time | none |
| 4 | Tensorlake | ≥ $0.2 [4] | ≥ $0.074 | ≥ $1,760 | Prepaid credit | 100 | ? | 2 vCPU / 4 GB | Active CPU | Active-CPU billing |
| 5 | Beam | ≥ $0.227 [5] | ≥ $0.227 | ≥ $2,088 (on Team) | Plan from $20/mo | 30 | ? | 2 vCPU / 4 GB | Running time | none |
| 6 | Modal | ≥ $0.238 [6] | ≥ $0.238 | ≥ $2,094 | Pay as you go | 100 | ? | 2 vCPU / 4 GB | Running time | none |

Not published, so not listed here: Novita AI, Google Cloud Agent Platform sandboxes, Islo, Runloop.

## Other pages

- [2 vCPU / 4 GB](https://roost.guide/agent-sandboxes/)
- [4 vCPU / 8 GB](https://roost.guide/agent-sandboxes/4-vcpu-8-gb/)
- [8 vCPU / 16 GB](https://roost.guide/agent-sandboxes/8-vcpu-16-gb/)
- [Devboxes](https://roost.guide/devboxes/)

## Filters

- [All](https://roost.guide/agent-sandboxes/)
- [Pay as you go](https://roost.guide/agent-sandboxes/pay-as-you-go/)
- [Prepaid](https://roost.guide/agent-sandboxes/prepaid/)
- [Monthly plan](https://roost.guide/agent-sandboxes/subscription/)
- [Free tier](https://roost.guide/agent-sandboxes/free-tier/)
- [100+ at once](https://roost.guide/agent-sandboxes/high-concurrency/)
- [Start limits published](https://roost.guide/agent-sandboxes/published-start-limits/)
- [No session limit](https://roost.guide/agent-sandboxes/no-session-limit/)
- [Active-CPU billing](https://roost.guide/agent-sandboxes/active-cpu-billing/)
- [Desktop](https://roost.guide/agent-sandboxes/desktop/)
- [Docker inside](https://roost.guide/agent-sandboxes/docker/)
- [SSH](https://roost.guide/agent-sandboxes/ssh/)
- [Snapshots](https://roost.guide/agent-sandboxes/snapshots/)
- [Fork](https://roost.guide/agent-sandboxes/fork/)
- [MicroVM](https://roost.guide/agent-sandboxes/microvm/)
- [Self-host](https://roost.guide/agent-sandboxes/self-hosted/)
- [Your cloud](https://roost.guide/agent-sandboxes/byoc/)
- [SOC 2](https://roost.guide/agent-sandboxes/soc2/)
- [Europe](https://roost.guide/agent-sandboxes/europe/)

## Sources

[1] https://www.koyeb.com/docs/reference/instances (checked 31 Oct 2026)
[2] https://northflank.com/pricing (checked 31 Oct 2026)
[3] https://www.daytona.io/pricing (checked 31 Oct 2026)
[4] https://www.tensorlake.ai/pricing (checked 31 Oct 2026)
[5] https://www.beam.cloud/pricing (checked 31 Oct 2026)
[6] https://modal.com/pricing (checked 31 Oct 2026)