Deploy a model
Deploy a model as a dedicated inference endpoint.
A deployed model runs as a dedicated inference endpoint on zCLOUD, the GPU cloud. Endpoints appear under **Inference** in the **Tasks** group of the zCLOUD menu. The zSTUDIO menu has no Inference item. This page creates the endpoint; to send it a request, see [Call your endpoint](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/call-your-endpoint).
Before you begin
- A model in [**Models**](https://console.zettabyte.com/library-model/list), either a preset from the catalog or one you brought yourself.
- GPU capacity for the endpoint's replicas. See [GPU instances](https://docs.zettabyte.com/docs/zcloud/compute/gpu-instances).
- To check that a model fits in the GPUs you will request, see [GPU memory calculator for LLMs](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/gpu-memory-calculator).
- An API key, which you will need to call the endpoint afterward. See [API keys](https://docs.zettabyte.com/docs/accounts-and-roles/api-keys).
Deploy a model
Open [**Inference**](https://console.zettabyte.com/task-inference/index) and select **Create Inference Task**. The form has an **Advanced Settings** switch; with it on, the form is the five-step wizard below. **Deploy** on a card of the **Pre-Trained Model** tab in **Models** also opens a **Create Inference Task** panel, at **Basic Information**.
- **Basic Information**: an **Endpoint Name** and a **Description**.
- **Environment Configuration**: a **Distributed Inference** toggle, an **Inference Engine** with **Preset Engine** (a preset such as `vLLM`) and **Custom Engine** options, the **Model**, and **Environment Variables**.
- **Instance Configuration**: an **Inference Mode** toggle (**Standard Mode** or **Multi-Role**), a **Computing Config** (**Resource Name**, GPU/CPU/Memory) with a **Complete Machine Scheduling** toggle, a **Number of Copies** (replicas), a **Channel Affinity** switch, and **Storage**. Storage accepts a **Storage Volume** (PVC), an **Ephemeral Volume** (memory-backed, sized in GiB), and **S3** (access key, secret key, mount directory, region, and a path; an external AWS S3 URL is also accepted). Mount directories must be unique. With **Channel Affinity** on, each replica registers a channel on the gateway side, and requests from the same client session prefer the last successful channel.
- **Advanced**: optional **Scaling**. With scaling on, choose **Timed Scaling** (target replicas on a daily schedule) or **Automatic expansion and contraction** (target replicas plus a **Scalability Index** of **GPU Utilization Rate** with a target %).
- **Overview**: review and create.
Endpoint status
In the Inference list each endpoint shows Name/ID, Model Name, Inference Engine, Status, Scaling (ready/target replicas), Creation Time, GPU Count, and Resource Name.
To confirm the endpoint is running, open **Inference** and check that its **Scaling** shows every replica running, for example `1/1`. **Scaling** compares the running replicas with the target **Number of Copies**.
You then call the endpoint over the inference API. The `Service-Name` header binds each request to your deployment and is built as `TENANT_ID.JOB_ALIAS_NAME`. It is a zCLOUD-specific (non-standard) header, so add it explicitly in every client. See [Call your endpoint](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/call-your-endpoint).
Next steps
- [Call your endpoint](https://docs.zettabyte.com/docs/zcloud/deploy-and-serve/call-your-endpoint): paste-ready cURL and Python examples for the running endpoint.
- [API keys](https://docs.zettabyte.com/docs/accounts-and-roles/api-keys): create and manage keys.