Skip to main content
The InferCrane provider connects the Vercel AI SDK to InferCrane’s OpenAI-compatible model API.

Install

Configure the API key in a trusted server environment:

Stream text

The adapter includes final token usage in streaming requests so applications can reconcile usage with InferCrane receipts.

Control reasoning effort

Qwen3.8 supports none, low, medium, and xhigh reasoning effort:

Connect a private endpoint

Use this form for a private canary, a dedicated deployment, or a regional InferCrane endpoint. Keep the credential in server code and never expose it in a browser bundle.

Direct compatibility without the package

The standard AI SDK OpenAI-compatible adapter can call the same endpoint:
The dedicated package provides the stable default endpoint, model identifiers, structured-output declaration, and InferCrane environment conventions.