Open Brain
VercelEngineering PostOfficial Source

AI Gateway now supports Private Inference

AI Gateway can route private model endpoints through the same APIs, agent connections, logs, and reporting as catalog models, with optional catalog-model fallback.

Vercel · Nov 16, 2026
Open Source Open MarkdownOpen JSON
Source Summary

Private Inference connects self-hosted, fine-tuned, or prerelease model endpoints to AI Gateway. Private models use the same authentication, SDKs, and APIs as catalog models; requests appear in shared logs and reporting. **Vercel CLI 65.16.0 or later** can register providers and models. If you run a private model behind a coding agent, this can consolidate its credentials, calls, and observability with catalog-model traffic. The gateway translates supported request formats, so switching models can be a model-ID change; an optional catalog-model fallback is available. Private models are scoped to your Vercel team and excluded from the public catalog. Check the documentation for supported model types, requirements, and billing details before connecting an endpoint.

Practical Implication

No separate implication has been recorded yet.

Agent-Ready Context
Private Inference connects self-hosted, fine-tuned, or prerelease model endpoints to AI Gateway. Private models use the same authentication, SDKs, and APIs as catalog models; requests appear in shared logs and reporting. **Vercel CLI 65.16.0 or later** can register providers and models.
If you run a private model behind a coding agent, this can consolidate its credentials, calls, and observability with catalog-model traffic. The gateway translates supported request formats, so switching models can be a model-ID change; an optional catalog-model fallback is available.
Private models are scoped to your Vercel team and excluded from the public catalog. Check the documentation for supported model types, requirements, and billing details before connecting an endpoint.
Connected Context · Feed7 Judgment

Private Inference extends AI Gateway’s shared authentication, APIs, logs, and reporting to team-scoped private endpoints, with optional catalog fallback. This makes gateway-based agent setups usable for private or prerelease models while preserving common observability. The CLI version requirement and endpoint compatibility checks are practical prerequisites before connecting one.

Context Map
infracoding#coding-agents#gateways#observability