AI Gateway adds virtual models: swap the model and settings without redeploying apps or reconfiguring agents
Synopsis
AI Gateway now offers virtual models: a saved configuration naming which model to use and with which settings, called by apps and agents through a custom model ID such as vmc/coding-agent. Previously, changing the model or settings meant redeploying apps or reconfiguring agents; now the change happens centrally in AI Gateway. It can also supply options a coding agent does not expose, such as reasoning effort, fast mode and prompt caching, and carry provider choice, fallback models and Zero Data Retention requirements.
Interpretation
Apps and agents can call a custom model ID instead of a provider model ID, and the gateway resolves it to a saved combination of model and settings. Changing the model or settings previously required redeploying apps or reconfiguring agents; those changes now happen centrally in AI Gateway. The capability is available on AI Gateway, works with every model type on the gateway, and is called as vmc/<slug>, for example vmc/coding-agent.
The saved configuration can carry runtime options that an agent does not itself expose, as well as routing and compliance requirements. Coding agents do not expose options such as reasoning effort, fast mode and prompt caching, so the virtual model sets them; provider choice, fallback models and Zero Data Retention are specified in the same configuration. The published creation example puts anthropic/claude-opus-5.5 together with fast mode and Zero Data Retention in one configuration for a reusable coding-agent setup.
Every request through a virtual model can carry custom reporting tags for attributing spend. Spend attribution can therefore sit at the virtual-model layer, listed alongside configuring coding agents and controlling routing and compliance as the three uses of virtual models. This use is listed alongside configuring coding agents and controlling routing and compliance as the three uses of virtual models.
Perspective
The demonstrated scope is virtual models on AI Gateway: created through the web page or the Vercel CLI, called as vmc/<slug>, and working with every model type on the gateway. The intended users are developers running apps and agents on that gateway.
The material does not say how virtual models affect requests already in flight when the underlying model or settings change, and gives no quantitative results; the value is presented only as a feature description.