Vercel AI SDK Adds Support for Grok 4.7 Model ID in xAI Provider Package

Vercel shipped version 2.0.95 of its @ai-sdk/xai provider package on September 21, adding a model identifier for "Grok 4.7" to the SDK's list of supported xAI models, according to the package's changelog on GitHub.

The release is a patch change, commit d3e1726, tagged with the message "feat(xai): add grok 4.7 model ID." Patch-level version bumps in the Vercel AI SDK typically indicate small, non-breaking additions — in this case, a new string identifier that developers can pass to the SDK's xai() provider function to route requests to the corresponding xAI model.

No other details accompanied the release. The changelog entry does not include context window size, pricing, benchmark scores, or a description of what distinguishes Grok 4.7 from prior Grok versions. Vercel's package releases document changes to the SDK itself, not the underlying model's capabilities — those specifications come from xAI directly, and xAI has not published accompanying documentation alongside this SDK update.

The @ai-sdk/xai package is one of several provider integrations maintained inside Vercel's open-source ai monorepo, which lets JavaScript and TypeScript developers call models from OpenAI, Anthropic, Google, Meta, xAI, and other providers through a single unified interface. When a provider like xAI ships a new model, Vercel typically updates its corresponding package with a new model ID string so developers can start making requests to it without waiting for a major SDK version change.

What this means: This release is infrastructure plumbing, not a model announcement. Vercel's changelog confirms that a model ID for "Grok 4.7" now exists inside the SDK's type definitions and request-routing logic, which signals that xAI's Grok 4.7 model is available for API access. But the actual specifications — context window, pricing, benchmarks, and what's new about it — have not been disclosed in this release. Developers using the Vercel AI SDK can now reference grok-4.7 in their code, but anyone evaluating the model itself should look for xAI's own announcement rather than relying on this package update for technical details.