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

# GPU Compute

> GPU compute options and specifications available on Airon's bare-metal infrastructure

<table>
  <thead>
    <tr>
      <th>GPU Model</th>
      <th>Configuration</th>
      <th>Price</th>
      <th>Memory</th>
      <th>Key Features</th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td><strong>H200 NVL PCI-E</strong></td>
      <td>4x or 8x GPUs</td>
      <td>\$2.70/hour</td>
      <td>-</td>
      <td>Single-node with NVLink</td>
    </tr>

    <tr>
      <td><strong>H100 NVL PCI-E</strong></td>
      <td>2x, 4x, or 8x GPUs</td>
      <td>\$1.88/hour</td>
      <td>-</td>
      <td>Single-node with NVLink</td>
    </tr>

    <tr>
      <td><strong>HGX H200</strong></td>
      <td>4x or 8x GPUs</td>
      <td>\$2.63/hour</td>
      <td>-</td>
      <td>Single and multi-node cluster</td>
    </tr>

    <tr>
      <td><strong>HGX B200</strong></td>
      <td>TBA</td>
      <td>\$4.13/hour</td>
      <td>-</td>
      <td>Blackwell GPU architecture</td>
    </tr>

    <tr>
      <td><strong>GB200</strong></td>
      <td>36 Grace CPUs + 72 Blackwell GPUs</td>
      <td>\$8.00/hour</td>
      <td>-</td>
      <td>Liquid-cooled rack</td>
    </tr>

    <tr>
      <td><strong>L40S</strong></td>
      <td>2x, 4x, 8x GPUs</td>
      <td>\$1.80/hour</td>
      <td>48GB</td>
      <td>PCIe configuration</td>
    </tr>

    <tr>
      <td><strong>RTX PRO 6000</strong></td>
      <td>1-8x GPUs</td>
      <td>\$1.24/hour</td>
      <td>-</td>
      <td>Blackwell Architecture</td>
    </tr>

    <tr>
      <td><strong>RTX A6000</strong></td>
      <td>1x or 2x GPUs</td>
      <td>\$0.60/hour</td>
      <td>48GB</td>
      <td>PCIe configuration</td>
    </tr>

    <tr>
      <td><strong>RTX A5000</strong></td>
      <td>1x or 2x GPUs</td>
      <td>\$0.45/hour</td>
      <td>24GB</td>
      <td>PCIe configuration</td>
    </tr>
  </tbody>
</table>

<Info>
  Pricing is shown as starting rates per hour. Actual pricing may vary based on configuration, region, and availability.
</Info>

## Getting Started

To deploy GPU compute resources:

1. **Choose your GPU configuration** based on your workload requirements
2. **Create a machine** using the Airon CLI or API
3. **Configure your environment** with the necessary frameworks and libraries
4. **Deploy your workload** and monitor performance

```bash
# Example: Create a machine with 4x H100 GPUs
airon machines create \
  --type gpu \
  --architecture x86 \
  --brand nvidia \
  --model h100 \
  --number 4 \
  --region us-west-2
```

For more detailed guidance on GPU compute deployment, see our [Getting Started guide](/compute/getting-started) or explore our [tutorials](/tutorials/tutorial).