# What changed in @ai-sdk/google-vertex > 57 changes in @ai-sdk/google-vertex 3 to 5 from its official release notes, upgrade guides and advisories: 5 breaking, 18 new APIs and options. Each one cited. Sources checked 2026-10-18. HTML: https://unstale.sh/npm/@ai-sdk/google-vertex/ To see which of these apply to a repository, run `npx @unstale/cli init` in it. ## @ai-sdk/google-vertex 5 ### Breaking changes - Forced tool use is unsupported: `required` and named tool choices fall back to `auto`, while `structuredOutputMode: 'jsonTool'` falls back to native structured outputs, each with a warning. (5.0.97) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.97) - Together AI Gemini image inputs are marked unsupported with the current single-image request mapping. (5.0.96) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.96) - Support for Imagen models and their integration was dropped following the API shutdown. (5.0.60) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.60) ### New APIs and options - Adds a provider option and provider metadata to resume responses that used up their maxOutputTokens budget. (4.0.213, 5.0.111) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.111), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.213) - Adds gemini-4-argon as a model id. (3.0.182, 4.0.213, 5.0.111) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.111), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.213), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.182) - Exposes the first request's prompt token count in provider metadata. (3.0.182, 4.0.213, 5.0.111) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.111), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.213), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.182) - Adds Claude Haiku 5.5 as a supported model, with a 128k maximum output token limit, adaptive thinking, and five effort levels. (5.0.105) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.105) - Adds the gemini-nano-banana-2.1 model ID for use with Google models. (5.0.103) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.103) - Adds experimental streaming lifecycle callbacks for speech generation and transcription telemetry. (5.0.100) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.100) - Adds the Claude Sonnet 5.5 model ID to @ai-sdk/google-vertex. (3.0.179, 4.0.208, 5.0.97) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.97), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) - Claude Sonnet 5.5 adds the `between_tools` thinking type. Disabled thinking and `reasoning: 'none'` map to it, budget-based thinking converts to adaptive thinking, and `xhigh` and `max` effort are lowered to `high` with a warning. (5.0.97) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.97) - Adds Claude Fable 5.1 model IDs to the Vertex provider. (5.0.84) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.84) - Adds the gemini-3.8-flash model for use with @ai-sdk/google-vertex. (3.0.173, 5.0.75) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.173), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.75) - Adds unary transcription via generateContent, with language detection, speaker diarization, word timestamps, and custom vocabulary. (5.0.66) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.66) - Adds streaming transcription over the Live API WebSocket, with VERBATIM or SMART transcription formatting. (5.0.66) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.66) ### Behavior and default changes - Requests that reach the per-request cap (CONTINUATION) are continued automatically, with usage summed across requests. (3.0.182, 5.0.111) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.111), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.182) - With the Interactions API, polls background: true model calls until they finish. (5.0.111) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.111) - Top-level embedding dimensions are mapped to provider settings, allowing providers to receive the requested dimensions. (5.0.109) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.109) - Thinking can be disabled up to high effort; xhigh and max are lowered to high with a warning when thinking is disabled. Token budgets are converted to adaptive thinking with a warning. (5.0.105) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.105) - Adds telemetry to speech generation and transcription, with provider usage propagation. (5.0.100) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.100) - Capability declarations now use confirmed model IDs, leaving unrecognized model names unknown; asynchronous lookups and middleware overrides can also resolve to unknown. (5.0.96) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.96) - Capability declarations now include single-image editing support for Together AI FLUX.2 Pro and Flex. (5.0.96) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.96) ### Fixes and other changes - Fixes surfacing errors that occur during streaming. (5.0.115) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.115) - Fixes hostname-part validation by using isValidHostnamePart. (5.0.100) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.100) - Uses a standards-compliant User-Agent header. (5.0.98) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.98) - Capability declarations now advertise file-input support for QuiverAI Arrow 2 and Arrow 2 Telos. (5.0.96) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.96) - Supported Vertex tool result URLs are now forwarded as function response file data. (5.0.95) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.95) - Corrects the advertised supported per-call limit for Gemini images. (5.0.87) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.87) - Fixes downloading tool result file URLs. (5.0.81) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.81) - Unclassified empty image results are retried, so these results no longer necessarily end an attempt without retrying. (5.0.77) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.77) - Google Vertex prompt blocks are marked as terminal. (5.0.77) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.77) - The default output token setting for Llama 4 was fixed. (5.0.76) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.76) - Vertex text embedding models now advertise a batch limit of 250. (5.0.70) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%405.0.70) ## @ai-sdk/google-vertex 4 ### Breaking changes - Google Vertex content-filtered results are treated as terminal rather than retryable. (4.0.198) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.198) ### New APIs and options - Adds the claude-haiku-5-5 model ID to @ai-sdk/google-vertex; the model is recognized with a 128k maximum output token limit and all five effort levels. (3.0.181, 4.0.211) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.211), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.181) - Adds the `between_tools` thinking type, the lowest thinking setting on `claude-sonnet-5-5`. (3.0.179, 4.0.208) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) - Adds the gemini-3.8-flash model to the Google provider. (4.0.197) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.197) - Adds unary transcription via generateContent, with language detection, speaker diarization, word timestamps, custom vocabulary, and VERBATIM or SMART formatting. (4.0.191) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.191) ### Behavior and default changes - Automatically continues requests that reach the per-request CONTINUATION cap, sums usage across requests, and exposes the first request's prompt token count. (4.0.213) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.213) - With the Interactions API, polls background model calls until they finish. (4.0.213) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.213) - For claude-haiku-5-5, an enabled thinking budget is converted to adaptive thinking with a warning because the model does not support thinking token budgets. (4.0.211) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.211) - For `claude-sonnet-5-5`, budget-based thinking is converted to adaptive thinking. (4.0.208) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208) - For `claude-sonnet-5-5`, `structuredOutputMode: 'jsonTool'` falls back to native structured outputs with a warning. (3.0.179, 4.0.208) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) - For `claude-sonnet-5-5`, disabled thinking is replaced with `between_tools` thinking, with a warning. (4.0.208) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208) - For `claude-sonnet-5-5`, `required` and named tool choices fall back to `auto` with a warning. (3.0.179, 4.0.208) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) ### Fixes and other changes - `xhigh` and `max` effort are lowered to `high` with a warning because the API rejects them with `between_tools`. (3.0.179, 4.0.208) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.208), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) - Forwards supported GCS tool result URLs as function response file data on Vertex Gemini 3 and later, limited to image, PDF, and text MIME types; HTTP(S) downloads remain supported on all Gemini generations. (4.0.206) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.206) - Unclassified empty image results are retried, with retry-attempt accounting preserved. (4.0.198) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%404.0.198) ## @ai-sdk/google-vertex 3 ### Breaking changes - For claude-sonnet-5-5, disabled thinking is replaced with between_tools thinking, with a warning. (3.0.179) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) ### New APIs and options - Adds the continuationToken provider option for resuming responses that used up their maxOutputTokens budget. (3.0.182) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.182) - A provider-independent classification of result retryability is added and preserved through the AI Gateway. (3.0.174) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.174) ### Behavior and default changes - For claude-sonnet-5-5, budget-based thinking is converted to adaptive thinking. (3.0.179) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.179) - Unclassified empty image results are retried, while Google Vertex RAI-filtered results are treated as terminal. (3.0.174) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.174) ### Fixes and other changes - Converts enabled thinking with a token budget to adaptive thinking and emits a warning because the model does not support thinking token budgets. (3.0.181) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.181) - Recognizes the model as having a 128k max output token limit and all five effort levels; when thinking is disabled, xhigh and max effort are lowered to high with a warning. (3.0.181) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.181) - Diagnostics from completed attempts are preserved. (3.0.174) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.174) - Google Vertex RAI-filtered results are marked as terminal. (3.0.174) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/google-vertex%403.0.174)