Vercel AI Gateway adds Virtual Models, stable model IDs you can repoint without redeploying

On October 28, 2026, Vercel added Virtual Models to AI Gateway. You create a task-named ID such as vmc/coding-agent in the dashboard or with vercel ai-gateway virtual-models create, choose the model behind it, and set reasoning effort, fast mode, prompt caching, provider order, fallback models, Zero Data Retention, and reporting tags. Apps pass the ID as the model string in the AI SDK or the OpenAI- and Anthropic-compatible APIs, and coding agents such as Claude Code and Codex can select it after vercel ai-gateway setup.

Editing the model behind an ID updates every caller within a few minutes, with no code change or redeploy. Virtual model settings override conflicting request settings, and Zero Data Retention and no-training rules are enforced even if a request tries to turn them off. A Virtual Models page shows requests, spend, and success rate per task. Vercel says AI Gateway passes through provider prices with no markup and no platform fee on inference.