Introducing Virtual Models on AI Gateway
By Steven Van ·
Teams can swap models, fallbacks and Zero Data Retention behind one ID in AI Gateway, with apps and coding agents picking changes up in minutes.
Vercel has introduced Virtual Models on AI Gateway, letting teams set a model and its settings behind a single ID instead of hard-coding the model string in an app. Vercel describes a virtual model as an abstraction of the model string and configuration that apps and developers use.
How it works
A virtual model ID is created for a task, such as vmc/coding-agent, using the Vercel CLI:
vercel ai-gateway virtual-models create coding-agent --model anthropic/claude-opus-5.5 --config '{"speed":"fast","zeroDataRetention":true}'
The ID is then passed as the model string wherever the task runs, for example as the model value in the AI SDK's streamText. Through the one ID, a team can change the model, configure fallbacks and priority, and enforce Zero Data Retention centrally.
Changing models without a redeploy
Switching models through AI Gateway was already a one-line change, but that line lives in application code, so it usually meant a redeploy. With a virtual model, the model is changed in AI Gateway instead, for example:
vercel ai-gateway virtual-models edit coding-agent --model openai/gpt-6.1-sol
Vercel says every app and agent that calls the ID picks up the new configuration within a few minutes, with no code change or deployment. The post notes that 125 new models from more than 10 labs became available on AI Gateway in the last 90 days, about 10 a week.
Coding agents
Coding agents such as Claude Code and Codex connect with vercel ai-gateway setup. The virtual model can then be selected in each harness's model picker or configuration. In Codex, a team's virtual models are listed in /model, and the slug also works with --model and in config.toml, as in codex --model vmc/coding-agent. Provider order, model fallbacks, regional inference and zero data retention can be applied to Codex requests this way.