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

# ASG GPU

> Dedicated GPU pods for AI workloads — Coming Soon

# ASG GPU

<Warning>
  **🔜 Coming Soon** — Dedicated GPU provisioning is on the roadmap. The tools described below are planned but not yet active. See [Tool Availability](/guide/tool-availability) for current status.
</Warning>

Rent dedicated GPU instances on-demand with per-second billing via MCP tool calls.

## Planned Capabilities

* **H100 80 GB, A100 80 GB, RTX 4090 24 GB** — Premium GPUs for every workload
* **Per-second billing** — No hourly minimums, pay only for what you use
* **Instant provisioning** — Target \<30 seconds from tool call to SSH
* **SSH access** — Full root control of your pod
* **Multi-GPU** — Scale to 8× GPU nodes for large training runs

## Planned Tools

| Tool            | Description         | Status         |
| :-------------- | :------------------ | :------------- |
| `gpu_provision` | Provision a GPU pod | 🔜 Coming Soon |
| `gpu_heartbeat` | Extend a lease      | 🔜 Coming Soon |
| `gpu_terminate` | Terminate a pod     | 🔜 Coming Soon |

## Available Today

While GPU provisioning is coming soon, you can use these tools right now:

* **[`inference_chat`](/services/inference)** — AI completions with 100+ models
* **[`sandbox_execute`](/services/sandbox)** — Isolated code execution
* **`optify_vram_estimate`** — Estimate VRAM requirements for models

## Pricing

GPU pricing will be announced when the service launches. See [Pricing](/billing/pricing).

<Info>
  **Interested in early access?** Contact us at [hello@asgcompute.com](mailto:hello@asgcompute.com) to join the GPU waitlist.
</Info>