# Blaxel vs Prime Sandboxes: big fleets compared

Blaxel and Prime both aim at very large fleets of sandboxes. On the 2 vCPU, 4 GB benchmark Blaxel costs $2.76 and Prime $1.52.

**Runtime charges 77% less than Blaxel and 58% less than Prime for the same
job:** 1,000 one-minute runs of a 2 vCPU, 4 GiB sandbox, each with 20
CPU-seconds of work. Blaxel prices memory alone, Prime prices every resource
while it runs, and Runtime prices the CPU the code used.

## At a glance

|            | Blaxel                                        | Prime Sandboxes                                | Runtime                                             |
| ---------- | --------------------------------------------- | ---------------------------------------------- | --------------------------------------------------- |
| Isolation  | Lightweight virtual machines                  | Hardware-virtualized microVM, own guest kernel | Firecracker microVM, own kernel                     |
| Billing    | $0.0000115 per GB-second active, CPU included | $0.02 per vCPU-hour, $0.0125 per GiB-hour      | $0.025 per measured vCPU-hour, $0.0075 per GiB-hour |
| Fleet      | Over 100,000 at once on the top tier          | 1,024 per account to start                     | 100 paid; 50 for a new account's first week         |
| Idle       | Standby when connections close; ~25 ms resume | Idle timeout off by default, up to 24 hours    | Pause or idle pause, memory kept 1 to 365 days      |
| Free start | Up to $200 of credit                          | None published                                 | 100 sandbox hours, no card                          |

## Standby against low rates

Blaxel makes idle sandboxes nearly free by dropping them to standby the moment
their connections close. Prime makes running sandboxes cheap per hour but
keeps billing until the sandbox ends or its optional idle timeout fires.
Runtime's idle pause, set with `idlePauseSeconds`, stops compute billing too,
and keeps memory for when the next request wakes it.

## Cost for the same job

The benchmark, 60 seconds a run (4 GB of active memory on Blaxel, a 5 GiB disk
on Prime):

```
Blaxel   Memory 1,000 × 60 s × 4 × $0.0000115                 = $2.76

Prime    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
         Total                                                  $1.52

Runtime  CPU    1,000 × 20 s / 3,600 × $0.025                = $0.14
         Memory 1,000 × 60 s / 3,600 × 4 × $0.0075           = $0.50
         Total                                                  $0.64
```

With both CPUs busy for the whole minute, the totals are $2.76 on Blaxel, $1.52 on Prime and $1.33 on Runtime.

## Which one to pick

- **Blaxel** for fleets of mostly idle sandboxes that resume in milliseconds.
- **Prime Sandboxes** for dense RL training runs.
- **Runtime** for the lowest cost while sandboxes run, and cheap paused
  storage.

See [Runtime vs Blaxel](/docs/blaxel-alternative) and
[Runtime vs Prime Sandboxes](/docs/prime-sandboxes-alternative).

## Sources

Blaxel rates checked 23 September 2026; Prime Sandboxes rates checked 25 September 2026.

- [Blaxel pricing](https://blaxel.ai/pricing), [Blaxel sandboxes](https://docs.blaxel.ai/Sandboxes/Overview)
- [Prime Sandboxes overview](https://docs.primeintellect.ai/sandboxes/overview), [Prime Sandboxes launch](https://www.primeintellect.ai/blog/sandboxes)
- Runtime [pricing](/docs/pricing)

Facts on this page were checked on 25 September 2026.