Sign in
← AI PulseSep 16, 2026

Wire · news · Multi-source brief

Vercel AI SDK Updates Include Experimental Evaluation Models and Tool Search

The Vercel AI SDK has been updated to version 7.0.104, introducing experimental evaluation model aliases, registry resolution, and enhanced tool search capabilities.

By Illumora Editorial

Source · Sep 16, 2026, 9:31 PM · On Illumora · Sep 16, 2026, 9:42 PM

Media from the primary source — shown here so you can stay on Illumora.

Synthesized from multiple allowlisted primaries on the same event. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →Vercel AI SDK Changelog — Release @ai-sdk/zai@3.0.13: Version Packages (#20860) · vercel/ai
Save

The Vercel AI SDK has released version 7.0.104, which includes several updates across its packages, such as @ai-sdk/zai@3.0.13, @ai-sdk/workflow-harness@1.0.114, @ai-sdk/xai@5.0.2, and @ai-sdk/vue@4.0.104. These updates introduce new features for model evaluation and tool integration.

Key Points

  • The Vercel AI SDK now includes experimental evaluation model aliases and registry resolution.
  • The customProvider function accepts evaluationModels for configuration.
  • Registries now expose evaluationModel for access.
  • The experimental_evaluate function accepts string IDs when an evaluation-capable default provider is explicitly configured.
  • Evaluation will not implicitly fall back to Gateway.
  • Model-resolution errors now identify evaluationModel.
  • The SDK supports tool search with direct tool calling.
  • A native tool search tool has been added.

Context

According to the Vercel AI SDK Changelog, these changes aim to provide more granular control over model evaluation and improve the integration of tools within applications. The updates ensure that stable provider contracts remain unchanged while introducing experimental features for developers to explore.

Why It Matters

These updates offer developers new ways to evaluate models and integrate external tools directly within their applications built with the Vercel AI SDK. The explicit configuration for evaluation models and the prevention of implicit fallback to Gateway provide more predictable behavior for model assessment.

What To Do

  • Review the Vercel AI SDK Changelog for ai@7.0.104 to understand the full scope of changes.
  • Test the experimental_evaluate function with string IDs if you are using an evaluation-capable default provider.
  • Explore the new tool search capabilities, including direct tool calling and the native tool search tool.
  • Note how customProvider now accepts evaluationModels and how registries expose evaluationModel.