Skip to main content

Quickstart

Deploy your first dedicated model endpoint and make your first API call. You will need an FPT AI Factory account with credits — go to Billing to top up before you start.

Step 1: Deploy a model​

  1. In the left sidebar, under Dedicated Inference, click Model Catalog.
  2. Browse the catalog and click Deploy on a model card. For a quick test, choose a small model (e.g. Mistral 7B).

Dedicated Inference — Model Catalog

  1. In the Deploy Dedicated Endpoint modal:
    • Enter the Hugging Face repo / URL (e.g. mistralai/Mistral-7B-Instruct-v0.3).
    • Select a Region.
    • Select Quick mode and choose a Size (Small / Medium / Large).
    • Review the ESTIMATED COST (PER HOUR / BIZ HOURS / 24/7).
  2. Click Deploy endpoint.

Deploy Dedicated Endpoint modal

See detailed guide: Deploy a Model

Step 2: Wait for Running status​

  1. In the left sidebar, under Dedicated Inference, click My endpoints.
  2. Find your new endpoint — its status shows Deploying.
  3. Wait until the status changes to Running.

My endpoints tab showing endpoint statuses

See detailed guide: Manage Endpoints

Step 3: Get an API key​

  1. Click the API keys tab in Dedicated Inference.
  2. Click + Create new key.
  3. Enter a Name for the key (e.g. quickstart-test).
  4. Click Generate Key, then copy the key — it is shown only once.

Create an API Key dialog

See detailed guide: Create an API Key

Step 4: Call the API​

Use the gateway URL from the endpoint detail page:

curl https://<ddi-gateway-domain>/v1/chat/completions \
  -H "Authorization: Bearer sk-vgw-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mistralai/Mistral-7B-Instruct-v0.3",
    "messages": [{"role": "user", "content": "Hello!"}],
    "max_tokens": 256
  }'

Find the gateway domain on the endpoint detail page. Get your API key from My endpoints → API keys.

See detailed guide: Call the Inference API

Step 5: Stop the endpoint when done​

To avoid unnecessary GPU charges, stop the endpoint when not in use:

  1. In the left sidebar, under Dedicated Inference, click My endpoints.
  2. Find your endpoint and click the Pause button in the Actions column.
  3. Confirm the action.

See detailed guide: Manage Endpoints