Vercel AI Gateway now supports virtual models. A virtual model is a saved configuration with a custom slug. Apps and agents call it by that slug instead of a provider model ID. It stores routing, fallbacks and compliance settings. Users can change the model or settings centrally without redeploying. Virtual models can configure coding agents with options like reasoning effort or fast mode. They also control routing, provider choice and compliance rules such as Zero Data Retention. Custom reporting tags can be attached to track spend. Virtual models can be created in the UI or via the Vercel CLI. They work with every model type on AI Gateway.
What changed
Apps and agents had to call specific model IDs directly with hard-coded settings.
What it unlocks
Changing model or settings for many apps or agents from one place without redeploying.
What you need to act on it
- an AI Gateway setup
- a virtual model created in AI Gateway
Sources