Quickstart: launch a GPU and deploy a model
Sign in, launch a GPU, and deploy a model to an inference endpoint.
These steps take you from a new sign-in to a running inference endpoint on zCLOUD, the GPU cloud. [How zCLOUD works](https://docs.zettabyte.com/docs/zcloud/introduction/how-zcloud-works) explains the terms used here, and [Call your endpoint](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/call-your-endpoint) covers sending requests to the endpoint.
Before you begin
- An account on console.zettabyte.com (username and password), set up as described in [Sign in and accounts](https://docs.zettabyte.com/docs/get-started/sign-in-and-accounts).
Steps
- Go to [console.zettabyte.com](https://console.zettabyte.com) and sign in. The zCLOUD workspace is named **AI Infrastructure Rental** in **Switch Platform**.
- Open [**Instance**](https://console.zettabyte.com/instance/list) and select **Launch Instance**.
- Check that **Resource Type** shows Standard GPU Instance, choose a **Computing Config** and set the **Count**.
- Check that **Purchase Cycle** shows On-Demand (Hourly), then select **Confirm**. What you'll see: a **Resources ready** notice, and the new instance in the list on the **Instance** page. If the form offers **Join The Queue** instead, see [The resource queue](https://docs.zettabyte.com/docs/zcloud/compute/resource-queue).
- Open [**Models**](https://console.zettabyte.com/library-model/list), choose a pre-trained model and select **Deploy** to open a step-by-step wizard. Deploy needs idle GPUs, which the instance from step 4 provides unless other work is using them.
- In the wizard, set the **Inference Engine** (for example `vLLM`), the **Computing Config**, the **Number of Copies** (replicas), and the storage, then complete the wizard. The fields are described in [Deploy a model](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/deploy-a-model). What you'll see: the endpoint under **Inference** in the zCLOUD **Tasks** menu, ready once its replicas are running (**Scaling** shows, for example, `1/1`). If the wizard shows "GPU resources are insufficient, please check.", no GPUs are idle.
Troubleshooting
If the endpoint does not become ready, see [Common issues](https://docs.zettabyte.com/docs/zcloud/troubleshooting/common-issues).
Next steps
- [Call your endpoint](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/call-your-endpoint): send requests to the running endpoint.
- [Stop work and release a GPU instance](https://docs.zettabyte.com/docs/zcloud/compute/stop-work-and-release): end the instance you launched in step 4 (**Terminate** in the console).
- [How billing works](https://docs.zettabyte.com/docs/billing/how-billing-works): wallet, payment, and consumption.