The Vercel AI SDK has updated its @ai-sdk/anthropic and @ai-sdk/gateway packages to include support for the Claude Opus 5.5 model. This integration was part of a release that also updated other dependencies within the SDK.
Key Points
- The Claude Opus 5.5 model ID has been added to the
@ai-sdk/anthropicpackage. - The model ID
anthropic/claude-opus-5.5is now supported in the@ai-sdk/gatewaypackage. - Models that consistently use adaptive thinking, such as claude-opus-5-5, claude-fable-5, and claude-fable-5-1, will no longer receive
thinking: { type: 'disabled' }or budget-based thinking settings. - The provider will issue a warning if unsupported thinking settings are passed to these models.
- Models that do not support forced tool use, including claude-opus-5-5 and claude-fable-5-1, will default to auto tool choice for required and named tool choices.
- These models will also revert to native structured outputs when
structuredOutputModeis specified.
Context
According to the Vercel AI SDK changelog, this update was part of a release initiated by the Changesets release GitHub action. The release included patch changes for ai@5.0.263, @ai-sdk/gateway@2.0.155, @ai-sdk/amazon-bedrock@3.0.133, @ai-sdk/anthropic@2.0.103, and @ai-sdk/angular@1.0.263.
Why It Matters
This update provides developers using the Vercel AI SDK with access to Claude Opus 5.5, expanding the range of available models and their capabilities. It also clarifies how certain advanced features, such as adaptive thinking and tool use, are handled for specific Anthropic models within the SDK.
What To Do
- Review the updated
@ai-sdk/anthropicand@ai-sdk/gatewaydocumentation for details on Claude Opus 5.5 integration. - Test applications with Claude Opus 5.5 to understand its behavior regarding adaptive thinking and tool use.
- Note the changes in how
thinkingandstructuredOutputModesettings are processed for models like claude-opus-5-5.