> ## Documentation Index
> Fetch the complete documentation index at: https://docs.darkbloom.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Pricing

> What inference, CI and agent sandboxes cost: every rate at a glance, from one shared prepaid balance.

Inference, CI and agent sandboxes draw on one prepaid Darkbloom balance. Add
funds at [console.darkbloom.dev](https://console.darkbloom.dev). These prices
are as of November 2026.

## CI

| Runner class | Platform | vCPU / memory | Per minute | Per hour |
| - | - | - | - | - |
| `linux-arm64-4cpu-16gb` | Linux | 4 vCPU / 16 GB | \$0.0025 | \$0.15 |
| `linux-arm64-8cpu-32gb` | Linux | 8 vCPU / 32 GB | \$0.005 | \$0.30 |
| `macos-arm64-8cpu-16gb` | macOS | 8 vCPU / 16 GB | \$0.02 | \$1.20 |
| `macos-arm64-12cpu-32gb` | macOS | 12 vCPU / 32 GB | \$0.04 | \$2.40 |

Billed per second, one-minute minimum, at most 60 minutes per job.

[CI pricing](https://docs.darkbloom.dev/ci/pricing.md) has worked examples and the rules on what is and is
not billed.

## Sandboxes

### Linux sizes

| Size | vCPU | Memory | Workspace | Per hour |
| - | - | - | - | - |
| `c4-m8` | 4 | 8 GiB | 32 GiB | \$0.192 |
| `c4-m16` | 4 | 16 GiB | 32 GiB | \$0.256 |
| `c8-m16` | 8 | 16 GiB | 64 GiB | \$0.384 |
| `c8-m32` | 8 | 32 GiB | 64 GiB | \$0.512 |
| `c12-m32` | 12 | 32 GiB | 64 GiB | \$0.640 |
| `c16-m32` | 16 | 32 GiB | 64 GiB | \$0.768 |
| `c16-m64` | 16 | 64 GiB | 64 GiB | \$1.024 |
| `c24-m48` | 24 | 48 GiB | 64 GiB | \$1.152 |
| `c24-m96` | 24 | 96 GiB | 64 GiB | \$1.536 |
| `c32-m64` | 32 | 64 GiB | 64 GiB | \$1.536 |
| `c32-m128` | 32 | 128 GiB | 64 GiB | \$2.048 |

Linux sizes follow this formula: $0.032 per vCPU-hour plus $0.008 per
GiB-hour of memory. For `c8-m32`, that is 8 × $0.032 plus 32 × $0.008, or
\$0.512 an hour.

### macOS sizes

| Size | vCPU | Memory | Workspace | Per hour |
| - | - | - | - | - |
| `mac-c4-m8` | 4 | 8 GiB | 32 GiB | \$0.35 |
| `mac-c4-m16` | 4 | 16 GiB | 32 GiB | \$0.45 |
| `mac-c8-m16` | 8 | 16 GiB | 64 GiB | \$0.70 |
| `mac-c8-m32` | 8 | 32 GiB | 64 GiB | \$0.90 |

Billed per second from ready, one-minute minimum per charge; stopped sandboxes are free.

[Sandbox pricing](https://docs.darkbloom.dev/sandboxes/pricing.md) has worked examples and the rules on what
is and is not billed.

## Inference

Inference is charged per token, at per-model rates in
[inference pricing](https://docs.darkbloom.dev/billing/pricing.md); the live prices are at
[api.darkbloom.dev/v1/pricing](https://api.darkbloom.dev/v1/pricing).

## Providers

Providers earn 90% of each CI and sandbox charge; see
[sandbox hosting](https://docs.darkbloom.dev/provider/sandbox-hosting.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.