Vercel Sandbox sandbox
Active CPU (I/O wait not billed) + provisioned memory (1-minute minimum) + per-creation fee. Rates for the default iad1 region. Each vCPU includes 2 GB memory.
Cost / code-agent task
$0.0092
2 vCPU / 4 GiB sandbox for 5 minutes at 10% CPU utilization (most time spent waiting on the LLM). Mirrors Vercel's own "AI code validation" example.
Free tier
Hobby: 5 active-CPU hours, 420 GB-hours, 5,000 creations per month. Personal/non-commercial use only per Vercel ToS.
source ↗ · 2026-10-17
Hobby plan is for personal or non-commercial use only (vercel.com/legal/terms).
Rates
| Item | Rate | Source | Fetched |
|---|---|---|---|
| Active CPU | $0.128 /vCPU-hour | vercel.com ↗ | 2026-10-17 |
| Provisioned memory | $0.0212 /GB-hour | vercel.com ↗ | 2026-10-17 |
| Sandbox creations | $0.6 /1M creations | vercel.com ↗ | 2026-10-17 |
| Snapshot storage | $0.08 /GB-month | vercel.com ↗ | 2026-10-17 |
Plans
| Plan | Monthly fee | Included | Concurrency | Max session | Source |
|---|---|---|---|---|---|
| Hobby | $0 | 5 CPU-h, 420 GB-h, 5,000 creations / mo | 10 | 45 minutes | source ↗ |
| Pro | $20 | $20/mo usage credit | 10,000 | 24 hours | source ↗ |
Limits & details
- Max vCPU / memory
- Hobby 4 / 8 GB; Pro 8 / 16 GB; Enterprise 32 / 64 GB ↗
- Disk
- 64 GB ephemeral NVMe ↗
- vCPU allocation rate
- Hobby 20–40/min; Pro up to 5,000/min ↗
- SDKs
- TypeScript, Python ↗
- Pricing page
- https://vercel.com/docs/vercel-sandbox/pricing
- Terms
- https://vercel.com/legal/terms
- Last verified
- 2026-10-17
Cold start
No Coldtally benchmark runs published yet. For broad speed coverage see ComputeSDK's open benchmarks (MIT-licensed data). Our methodology explains how we use them.
Changes
What changed in agent infra, weekly
Every pricing, plan and limit change we detect for sandboxes and cloud browsers, with sources. One email a week. No vendor-paid content.