Skip to main content

Using Models in Serverless Mode

Serverless mode lets you call models immediately, with no instance to create and no server to configure. Use the prebuilt endpoints provided by Elice AI Cloud, and pay only for what you use.

Characteristics​

  • Ready to use, no infrastructure to prepare
  • Per-call billing
  • Automatic adaptation to changing traffic
  • Suited to testing, prototyping, and flexible deployments

Using a Serverless Endpoint​

Selecting a model that supports Serverless in the Model Library makes its Serverless endpoint immediately available.

Use — Serverless

  1. ML API → Go to Model Library
  2. Choose a model that supports Serverless
  3. Review the API call information provided
  4. Call with your API key included

Calling a Serverless Endpoint​

Calls to a Serverless endpoint require API key authentication. Include the API key in the request header.

For API keys, see Managing API Keys.

For how to call deployed models, see API Requests.

Serverless Usage Monitoring​

The Serverless Usage page in ML API shows your usage and cost.

Serverless monitoring

What you can see​

  • Total spend
  • Calls per time period
  • Token, seconds (audio), megapixel (image) usage
  • Filters by model and by API key

This helps you analyze usage trends and manage costs efficiently.

Pricing​

Serverless is per-call billing, and pricing policy may vary by model type.

  • Billed by input/output tokens
  • Call volume shown in a chart
  • Use the date filter to view history