Vercel AI Gateway adds Private Inference for self-hosted and fine-tuned models
Vercel announced that AI Gateway now supports Private Inference, allowing users to connect their own model endpoints, such as fine-tuned, self-hosted, or prerelease models, to AI Gateway. These private models use the same authentication, routing, and observability as catalog models, are visible only to the Vercel team, and can be called by model ID in the form private/<slug>.

Why it matters
Teams can manage private models through the same keys, client code, and logs used for catalog models, instead of maintaining separate integrations. Private models can also be configured with fallbacks to catalog models if the private endpoint fails, and requests can be switched between private and catalog models by changing only the model ID.
Who should care
Enterprises and engineering teams running fine-tuned or self-hosted models, developers using AI Gateway, teams using coding agents such as Claude Code, Codex, or Cursor with AI Gateway.
What you can do
Add a private provider in the Vercel dashboard by entering your endpoint's base URL, API format, and credentials, then register its models and call them using the private/<slug> model ID.