Skip to main content
Version: 3.6.1-rc.2

Quickstart: deploy and share a model

This quickstart is for platform administrators. In about 15 minutes you deploy a model, give it a stable name with a model alias, and share that alias so people in your tenants can use it.

You need:

  • a platform administrator account for the CM portal
  • enough free GPU memory for the model. See VRAM requirements
  • for a gated Hugging Face model, a Hugging Face account and access token

1. Deploy a model​

This example deploys the small Qwen/Qwen3-0.6B model from Hugging Face.

  1. From the CM portal, select Manage platform at the bottom of the sidebar, then Models, and select Deploy model.
  2. Keep Preset set to No preset (manual configuration) and Model source set to HuggingFace. To use a ready-made configuration instead, pick a model from Preset.
  3. Enter Qwen/Qwen3-0.6B under Model URL, then fill in a unique API model id.
  4. Choose a Node and GPUs with enough free memory. For a gated model, enter a HuggingFace token.
  5. Select Deploy. The Models list shows the deployment's status while it starts. To check startup progress, open the model from the list and select the Logs tab.

2. Create a model alias​

A model alias is the stable name that clients send as the model. Later you can point it at another model, and those clients keep working without a change.

  1. From the CM portal, in Manage platform → Models, select the Model Aliases tab, then Create model alias.
  2. Enter a Public model ID, such as chat-default. This is the name clients send, and it can't be changed later.
  3. Set Target kind to Model deployment and choose your new deployment as Target model.
  4. Select Create alias. The alias's page opens.

3. Share the model or the alias​

On the alias's page or the model's page, select the Access tab:

  • Share with all tenants lets every user on the platform call it.
  • Shared with tenants limits it to the members of the tenants you select.

Select Save access.

4. Check it​

Members of a tenant you shared with now find the alias on the Model Aliases tab, or the model on the Deployments tab, of their tenant Models page. They can select it in Chat and grant an API key on it.

More resources​