Which is better, AWS Lambda MicroVMs or Prime Sandboxes?
It depends on where your stack lives and how many sandboxes run at the same time. For agents that mostly wait on a model, Runtime costs least: $0.64 for the benchmark, against $1.52 on Prime and $4.20 on Lambda, and it stays cheapest with both CPUs busy. Every pause keeps memory and processes. Pick Lambda MicroVMs when IAM, VPC access and AWS regions matter most; pick Prime Sandboxes for RL rollouts by the thousand, with 1,024 sandboxes per account to start.
At a glance
| AWS Lambda MicroVMs | Prime Sandboxes | Runtime | |
|---|---|---|---|
| Isolation | Firecracker microVM | Hardware-virtualized microVM, own guest kernel | Firecracker microVM, own kernel |
| CPU | $0.0997 per baseline vCPU-hour | $0.02 per vCPU-hour while running | $0.025 per vCPU-hour of measured CPU, with a small floor |
| Memory | $0.0132 per GB-hour | $0.0125 per GiB-hour | $0.0075 per reserved GiB-hour |
| Largest | A 4 vCPU, 8 GB baseline | 16 vCPUs, 64 GiB, 128 GiB of disk | 16 vCPUs and 64 GiB, paid |
| At once | Set by an account memory quota per Region | 1,024 per account to start | 100 on a paid account |
| Suspend | Memory and disk in a snapshot, up to 8 hours | Filesystem checkpoints; processes start fresh | Memory and processes while you have credit |
Why the gap is so wide
Memory costs about the same on both. CPU does not: Lambda charges five times Prime's rate per vCPU-hour, and both charge it for every second a sandbox runs. Two vCPUs for a minute cost Lambda 0.33 cents and Prime 0.07 cents before memory. Runtime's measured CPU is dearer per hour than Prime's but is only charged while the code computes.
Cost for the same job
1,000 runs of 60 seconds, 2 vCPU and 4 GiB (4 GB on Lambda), 20 CPU-seconds each, and Prime's 5 GiB default disk:
- CPU
- 1,000 × 60 s × 2 × $0.0000276944
- $3.32
- Memory
- 1,000 × 60 s × 4 × $0.0000036667
- $0.88
- 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
A month of 100,000 runs: $420.33 on Lambda, $151.67 on Prime, $63.89 on Runtime. Kept busy for the whole minute on both CPUs, the job is $4.20 on Lambda, $1.52 on Prime and $1.33 on Runtime.
Which one to pick
- Lambda MicroVMs where IAM, VPC access and AWS regions matter most.
- Prime Sandboxes for RL rollouts by the thousand on Prime Intellect.
- Runtime for the lowest bill at any load, and sandboxes that pause with their memory and fork with it.
More: Runtime vs AWS Lambda MicroVMs, Runtime vs Prime Sandboxes.
Runtime Cloud
Runtime is cheaper than both: 85% under Lambda and 58% under Prime. The benchmark, 1,000 one-minute runs at 2 vCPU and 4 GiB doing 20 CPU-seconds of work each, costs $4.20 on AWS Lambda MicroVMs, $1.52 on Prime Sandboxes and $0.64 on Runtime.
Sign in with npx withruntime login and run the benchmark on the free monthly usage. The migration guide covers moving a project that already runs elsewhere.
Sources
Lambda and Prime rates checked 25 October 2026.
- AWS Lambda pricing, Lambda MicroVMs
- Prime Sandboxes overview, Prime Sandboxes launch
- Runtime pricing
Facts on this page were checked on 26 October 2026.