Which is better, Vercel Sandbox or Prime Sandboxes?
It depends on where the code is run from and how much of it waits. For agents that mostly wait on a model, Runtime costs $0.64 on the benchmark, against $1.52 on Prime and $2.12 on Vercel. Runtime also has a pause that keeps memory, and a one-import switch from Vercel's SDK. Pick Vercel Sandbox for code run from a Vercel app, close to its users. Pick Prime Sandboxes for RL rollouts by the thousand on Prime's platform.
At a glance
| Vercel Sandbox | Prime Sandboxes | Runtime | |
|---|---|---|---|
| Isolation | Firecracker microVM, own kernel | Hardware-virtualized microVM, own guest kernel | Firecracker microVM, own kernel |
| CPU | $0.128 per active CPU-hour, iad1 |
$0.02 per vCPU-hour while running | $0.025 per vCPU-hour of measured CPU, with a small floor |
| Memory | $0.0212 per provisioned GB-hour | $0.0125 per GiB-hour | $0.0075 per reserved GiB-hour |
| Built for | Code from apps deployed on Vercel | Agentic RL training on Prime Intellect | Agents of any kind, by API, CLI, SDK or MCP |
| Snapshots | Filesystem snapshots and forks | Filesystem checkpoints and forks; no memory | Memory, processes and files |
| Price terms | Hobby allowance, then Pro | Published through 22 December 2026 | Fixed in each quote; no plan |
Two meters, one lesson
Prime's CPU rate is about a sixth of Vercel's, but Prime charges it for every second, working or not. Vercel charges only the seconds that work. On the job below, busy a sixth of the time, Prime still comes out 61 cents cheaper, and on a busy one it wins by more. Runtime combines Vercel's approach, billing only active CPU, with a rate close to Prime's.
Cost for the same job
2 vCPU, 4 GiB (4 GB on Vercel), 60 seconds, 20 CPU-seconds, 1,000 runs, with Prime's 5 GiB default disk:
- CPU
- 1,000 × 20 s / 3,600 × $0.128
- $0.71
- Memory
- 1,000 × 60 s / 3,600 × 4 × $0.0212
- $1.41
- CPU
- 1,000 × 60 s / 3,600 × 2 × $0.02
- $0.67
- Memory
- 1,000 × 60 s / 3,600 × 4 × $0.0125
- $0.83
- Disk
- 1,000 × 60 s / 3,600 × 5 × $0.0002
- $0.02
- CPU
- 1,000 × 20 s / 3,600 × $0.025
- $0.14
- Memory
- 1,000 × 60 s / 3,600 × 4 × $0.0075
- $0.50
At 100,000 runs a month: $212.44 on Vercel, $151.67 on Prime, $63.89 on Runtime. With both CPUs busy for the whole minute, the job is $5.68 on Vercel, $1.52 on Prime and $1.33 on Runtime.
Which one to pick
- Vercel Sandbox for code run from a Vercel app, close to its users.
- Prime Sandboxes for RL rollouts by the thousand on Prime's platform.
- Runtime for the lowest price at any load, memory-keeping pause and forks, and one-import switching from Vercel's SDK.
See Runtime vs Vercel Sandbox and Runtime vs Prime Sandboxes.
Runtime Cloud
Runtime costs 70% less than Vercel and 58% less than Prime on the benchmark. A thousand one-minute runs of a 2 vCPU, 4 GiB sandbox, 20 CPU-seconds each, are $2.12 on Vercel Sandbox, $1.52 on Prime Sandboxes and $0.64 on Runtime.
Use the free usage after npx withruntime login. To move Vercel Sandbox or
Prime code, read the migration guide.
Sources
Vercel rates checked 25 October 2026; Prime rates checked 25 October 2026.
- Vercel Sandbox pricing, Understanding Vercel Sandboxes
- Prime Sandboxes overview, Prime Sandboxes launch
- Runtime pricing
Facts on this page were checked on 26 October 2026.